Benchmarks

Arena · WorkSep 9, 2026Study

Cost / quality frontier

13 models · score and cost

Claude Fable 5.1 (Max)14.73 score

Anthropiccost 6.93

Claude Opus 5 (High)11.27 score

Anthropiccost 3.91

Claude Fable 5 (High)9.58 score

Anthropiccost 3.34

GPT 5.6 Sol (xHigh)8.35 score

OpenAIcost 1.72

Claude Sonnet 5 (High)6.66 score

Anthropiccost 1.39

Kimi K3 (Max)6.66 score

Moonshotcost 1.06

GLM 5.2 (Max)5.54 score

Z.aicost 0.98

Hy4 preview5.49 score

Tencentcost 0.25

Deepseek V4 Flash (High) (20260731)2.84 score

DeepSeekcost 0.16

GLM 5.3 Flash1.8 score

Z.aicost 0.09

GPT 5.6 Luna (xHigh)1.06 score

OpenAIcost 0.07

Hy3-4.36 score

Tencentcost 0.04

Mimo V2.5 Pro-5.5 score

Xiaomicost 0.04

Arena · Work · published resultsRead original study ↗

SourceJun 8, 2026Study

On-policy distillation on GDPval

Distilled vs. base · 220 tasks per model

Distilled winsTiesBase wins

Nemotron 3 Nano+20.9 pp net win–loss

1232077

55.9% wins+5.1 weighted rubric pointsdistilled from Nemotron 3 Super

Qwen3.5-9B+21.4 pp net win–loss

1261579

57.3% wins+7.9 weighted rubric pointsdistilled from Qwen3.5-397B-A17B

Source · published resultsRead original study ↗

More comparisons

Website · 4 outputs

Build a playable 3D flying game in the browser (Three.js): pilot a paper plane through floating islands at golden hour. Soft volumetric clouds, waterfalls spilling off the islands, flocks of birds, glowing rings to fly through, wind trails, a cinematic intro fly-by and a score. Hand-painted storybook look, gorgeous from the first frame.

Inspect the task →

Website · 3 outputs

Build a playable 3D hover-bike racing game in the browser (Three.js). Neon cyberpunk city at night, rain, glowing billboards, bloom. Race through traffic with boost pads, ramps and near-miss sparks; speedometer, lap timer and a dramatic crash + replay screen. Make it look like a premium console game trailer the moment it loads.

Inspect the task →

Other · 4 outputs

Create a 12-second cinematic 3D product reveal film in the browser (Three.js) for a fictional smartwatch called ORBIT. Dark studio, dramatic rim lighting, slow camera orbit and dolly, liquid-metal reflections, a particle burst on the reveal, bold kinetic typography, logo end card. Auto-plays and loops like an Apple keynote launch film.

Inspect the task →

Website · 4 outputs

Build a playable 3D monster-truck smash game in the browser (Three.js + physics). Drive a giant monster truck through a colourful toy city and smash everything: towers of blocks topple, cars flip, glass shatters into debris with real physics, slow-motion on big crashes, dust and sparks, a destruction score. Should feel explosive and satisfying from the first second.

Inspect the task →

Loading answer…

Other · 3 outputs

Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.

Inspect the task →

Video · 2 outputs

https://www.instagram.com/reel/Dd4VIgTq8e7/ reference this video! you must reference the asset of the video. music, sound effect etc... and what u need to describe is https://evals.ag/ site!

Inspect the task →

Video · 2 outputs

Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.

Inspect the task →

Video · 1 output

Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.

Inspect the task →

Video · 1 output

make a demo video about https://evals.ag/

Inspect the task →

Video · 4 outputs

reference this video for motion graphic demo https://x.com/samgrows/status/2107571800020222393?s=20 you must describe this site https://evals.ag/

Inspect the task →

Video · 1 output

Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.

Inspect the task →

Video · 1 output

reference this video for motion graphic demo https://x.com/samgrows/status/2107571800020222393?s=20 you must describe this site https://evals.ag/

Inspect the task →

Video · 1 output

could you make a demo video reference this site https://x.com/jordanarchivess/status/2107536262894915762?s=20. what you need to make it site is https://evals.ag. you must use tts with elevenlabs v4 api

Inspect the task →

Video · 1 output

START NOW. Do not ask me any questions and do not wait for input — every input is below. Make every unspecified decision yourself, log it in DECISIONS.md, and keep going until ONE finished video file exists. The only deliverable is that one .mp4. Task Make ONE ~60s landscape demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. Output: /Users/evan/workspace/video-generator/videos/evals-demo3/renders/evals-demo3.mp4, 1920×1080, 30fps, H.264 + AAC 192k, −14 LUFS. Nothing else to deliver (no vertical version, no variants). Work dir: repo /Users/evan/workspace/video-generator (Claude Code on this Mac). Build in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-demo3/. Render: npx --yes hyperframes@0.8.101 render -o <out> --quiet from the project dir. Use curl for HTTP/API calls (python has no SSL certs). ElevenLabs key: ELEVENLABS_API_KEY in /Users/evan/workspace/video-generator/.env. Story (English voiceover + on-screen type) Hook, first 3s must be instantly clear: real AI-made GAME/VIDEO outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." Type a real game/video task into the Evals composer → Battle → up to 4 setups (Model · Skill · Environment, each landing with real values and moving UI). Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. Blind compare: two moving outputs, "Which would you choose?" → vote ("I prefer A"). Run record of that same output (real LLM, skill, environment, turn time, tokens). Setup leaderboard ("Setups, not just models."). Community wall of real game/video/shorts outputs. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. Content rules Show ONLY videos, games, 3D scenes, motion videos, shorts made in Evals. No websites / landing pages. Main thread = the real 4-setup Sonic battle: prompt "Build a polished playable Sonic-style 3D platform game for the browser.", clips /Users/evan/workspace/video-generator/videos/evals-launch/assets/clips/run_a.mp4 … run_d.mp4, run https://evals-app.vercel.app/runs/run_597Q2CRE4FXNJ2R7HZXMVF4Z03. Capture fresh real UI from https://evals-app.vercel.app with Playwright at 2× DPR: composer, Run-mode menu, setups stepper, compare page, output run record, Leaderboards, Explore → Outputs. If a page needs sign-in, use the public pages. Record extra community game/video outputs from their public "Open preview" links. Screencast 1600×900, 30fps, 6–7s, with keys pressed or orbit so they move. Reuse the pattern in /Users/evan/workspace/video-generator/videos/evals-launch/scripts/rec.mjs. No real-person likeness, no invented features/numbers/names. Only show model names you see in captures (e.g. Claude Opus 5.5, GPT-6 Astra). Voiceover (must flow continuously) ElevenLabs voice Will bIHbv24MWmeRgasZH58o, model eleven_v4, stability 0.15, similarity 0.8, style 0.8, speed 1.1, with emotion tags in the text ([excited] [confident] [curious] [emphatic] [quick]). Use the /with-timestamps endpoint. Write ~165–180 words so the voice fills the whole film. Generate it as ONE take and lay it unspliced — never cut it into phrases or time-stretch pieces. Longest silence inside the VO ≤ 0.35s. No clipped words. The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. Verify with Whisper that every word is there and no tags are spoken. If "Evals" sounds like "evils", respell it "E-vals". Music & sound New modern premium BGM, not cheesy. Generate 2–3 candidates with ElevenLabs Music: "modern minimal tech launch, punchy kick, deep 808, crisp hats, glassy synth plucks, Apple/Linear keynote feel, no cheesy EDM risers, no corporate/stock vibe, 120 BPM". Pick the cleanest; detect BPM and the first downbeat. VO leads at about −16 LUFS. Music ducks about 6dB only while he speaks (smooth envelope). SFX bus ~3dB under the music, above it only on hits. At 4–6 big hits, layer reverse swell + boom (HP 60Hz) + transient + glitch. Light ticks and whooshes elsewhere — never on every cut. Look Fresh premium design system, written once in DESIGN.md: one sans font, one accent colour real UI captures as rounded cards (16–22px) with soft shadows dark/light section switches Never: outlines/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups. Motion (non-negotiable — this is the feel) Reference for feel: /Users/evan/workspace/video-generator/videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4; rules in /Users/evan/workspace/video-generator/videos/evals-hype/MOTION_FIX.md. One camera = a pure function of time with velocity-continuous easing. Deterministic: no timers or randomness. Two easing families only: snappy ≈25% of remaining distance per frame, floaty ≈12%. Exits accelerate into the cut; the next shot carries the velocity through it. Real whip-pans: both shots travel together over 0.3–0.45s with velocity blur. Overlays are screen-locked. Overlapping action, no start-stop chains. Every hold drifts 1–3%. No dead stops, no frozen frames, no double cuts. Word-by-word reveals 2–4 frames apart; the newest word flashes the accent. Cuts on the beat ±1 frame. Speed by re-rendering: render at --fps 30/speedup (e.g. --fps 95/4 = 1.263×), play back at 30fps. Impact FX in post (ffmpeg) on 4–6 big hits only: +8.5% zoom punch (τ 0.09s) 16px shake at ~22Hz (τ 0.1s) rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash A lighter version on secondary cuts. QA (do it, then deliver) python3 /Users/evan/workspace/video-generator/videos/evals-hype/scripts/motion.py <render> <bpm> <first_beat>: no dead stops, no frozen holds, cuts on the beat. sh /Users/evan/workspace/video-generator/videos/evals-hype/scripts/strip.sh <render> <t> <out.png> at every transition — look at each strip. VO RMS gap check (≤ 0.35s) plus a Whisper pass on the final mix. Text readable and inside the margins; every name and number on screen is real. Fix anything that fails, re-render, and finish with the single file videos/evals-demo3/renders/evals-demo3.mp4.

Inspect the task →

Other · 2 outputs

Make a ~60s product launch film for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. 16:9, 1920×1080, 30fps, music-locked, no narration — on-screen type tells the story. Build it in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-<name>/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use my logged-in Chrome session for anything behind sign-in). Motion reference: videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its feel exactly; change only scenario, design and assets. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Download the reference and study it frame by frame. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. strip.sh (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver renders/evals-<name>-v1.mp4 plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.

Inspect the task →

Video · 2 outputs

Make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464 You must reference this site https://evals.ag/. You must use font sanserif. Motion must be speedy. Never make ai slop.

Inspect the task →

Video · 2 outputs

Make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464 You must reference this site https://evals.ag/. You must use font sanserif. Motion must be speedy. Never make ai slop.

Inspect the task →

Video · 1 output

Evals landscape demo-video prompt (approved "감도") Copy everything below the line into a new session. Make a ~60s landscape product demo video for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. Horizontal 16:9, 1920×1080, 30fps (for YouTube / website / X timeline). Not a vertical Short/Reel — no 9:16 output, no vertical crops, no phone safe-zones. Music-locked, no narration — on-screen type tells the story while the real product is demoed end to end. Build it in HyperFrames (HTML + one paused GSAP timeline) under /Users/evan/workspace/video-generator/videos/evals-demo3/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use the logged-in Claude in Chrome session if connected; otherwise public pages). Motion reference (local file, already downloaded): /Users/evan/workspace/video-generator/videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its motion feel exactly (it is also 16:9); change only scenario, design and assets. Its 26s cut-down is just the reference for feel — this video runs the full ~60s. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." Run it autonomously — do not ask me for input Work in Claude Code with the repo /Users/evan/workspace/video-generator as the working directory. Every path below is absolute; everything you need is already on this machine. Do NOT stop to ask questions or request inputs. Every input is specified here; for anything not specified, pick a sensible default, write the decision into DECISIONS.md, and keep going until the final file is rendered. If something is unavailable, use the fallback and continue: Chrome not connected or a page needs sign-in → use the public pages (https://evals-app.vercel.app home, /compare, leaderboards, explore) with Playwright, and record outputs from their public "Open preview" links. Can't run a new Battle → use public outputs that share the same prompt (the /compare pages show pairs). Music: pick an unused track from /Users/evan/workspace/video-generator/videos/evals-hype/music3/ or music2/. SFX: /Users/evan/workspace/video-generator/videos/evals-hype/sfx2/, ref2/sfx3/, ref2/sfx4/. QA scripts: /Users/evan/workspace/video-generator/videos/evals-hype/scripts/motion.py and strip.sh (also beats.py for BPM). If they fail, write equivalents. Render: npx --yes hyperframes@0.8.101 render -o <out.mp4> --quiet from the project dir (--fps 95/4 etc. for speed re-render). API calls via curl (python has no SSL certs here). What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Study the local reference file frame by frame (ffmpeg frame dumps + frame-diff). Its design notes are in /Users/evan/workspace/video-generator/videos/evals-hype/ref2/SPEC.md and MOTION_FIX.md — read them, but design and copy must be new. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) python3 /Users/evan/workspace/video-generator/videos/evals-hype/scripts/motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. sh /Users/evan/workspace/video-generator/videos/evals-hype/scripts/strip.sh <render> <t> <out.png> (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver one 1920×1080 file renders/evals-demo3-v1.mp4 (no vertical version) plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.

Inspect the task →

$Make a ~60s product launch film for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. 16:9, 1920×1080, 30fps, music-locked, no narration — on-screen type tells the story. Build it in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-<name>/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use my logged-in Chrome session for anything behind sign-in). Motion reference: videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its feel exactly; change only scenario, design and assets. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Download the reference and study it frame by frame. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. strip.sh (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver renders/evals-<name>-v1.mp4 plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.$ Other · 1 output

Make a ~60s product launch film for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. 16:9, 1920×1080, 30fps, music-locked, no narration — on-screen type tells the story. Build it in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-<name>/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use my logged-in Chrome session for anything behind sign-in). Motion reference: videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its feel exactly; change only scenario, design and assets. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Download the reference and study it frame by frame. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. strip.sh (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver renders/evals-<name>-v1.mp4 plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.

Inspect the task →

Video · 2 outputs

Make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464 You must reference this site https://evals.ag/. You must use font sanserif. Motion must be speedy. Never make ai slop.

Inspect the task →

Video · 2 outputs

Make me a 15-second trailer for an F1 movie starring Ariana Grande

Inspect the task →

Video · 2 outputs

내가 지금 레퍼런스로 넣은 이미지들의 무드와 비슷하게 cozy한느낌의 브랜드 필름을 만들어줘

Inspect the task →

Video · 1 output

make a viral video in tiktok trending!

Inspect the task →

Video · 2 outputs

make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464?s=20 what you make it is https://evals.ag/ site

Inspect the task →

Website · 2 outputs

Build a single-page landing site for a coffee subscription startup with a hero, pricing table and signup form.

Inspect the task →

Video · 2 outputs

make a demo intro video reference this video https://x.com/rajsinghfirst/status/2107161801322549725?s=20 demo site is https://evals.ag ! this site

Inspect the task →

Video · 2 outputs

make a viral video reference this thing https://x.com/motion_conquest/status/2106786068712558848?s=20 what you need to make it is https://evals.ag/ site! demo vide

Inspect the task →

Video · 2 outputs

make a demo intro video reference this video https://x.com/rajsinghfirst/status/2107161801322549725?s=20 demo site is https://evals.ag ! this site

Inspect the task →

Other · 2 outputs

make a ai influencer for x viral

Inspect the task →

Video · 2 outputs

make a viral video in x

Inspect the task →

Website · 2 outputs

Build a landing page for a neighborhood bakery: menu, opening hours, map and a reservation button. It should look great on phones.

Inspect the task →

Website · 1 output

make a 2d kawai samurai game.

Inspect the task →

Website · 4 outputs

Build a polished playable Sonic-style 3D platform game for the browser. Include a fast character, rings to collect, ramps, jumping, hazards, checkpoints, score and restart. Make speed and platforming feel satisfying. Deliver a working browser artifact.

Inspect the task →

Website · 2 outputs

Build a polished playable Sonic-style 3D platform game for the browser. Include a fast character, rings to collect, ramps, jumping, hazards, checkpoints, score and restart. Make speed and platforming feel satisfying. Deliver a working browser artifact.

Inspect the task →

Document · 2 outputs

1+1?

Inspect the task →

Video · 2 outputs

Make me a 15-second ad with MapleStory characters built in 3D

Inspect the task →

Website · 2 outputs

Make MapleStory as a 3D game I can play in the browser

Inspect the task →

Video · 2 outputs

I want the whole human story in one shot — from the first people to the AI we use now. 15 seconds, 16:9, no narration, no on-screen text until the very last frame. One continuous move, never cutting. The camera pushes forward and the world changes around it: firelight on a cave wall, a hand pressed in ochre, then fields and the first walls, then a printing press, then steel and smoke, then a wall of city light, then a room of humming machines, then the light of a screen on a face in the dark. Each era arrives while the last one is still leaving — the fire becomes a furnace, the cave painting becomes a page, the page becomes a screen. Hand things off on shape: something round stays round, something bright stays in the same corner of the frame. Keep a human in almost every era, always small in the frame, so the scale of the thing reads against a person. No famous faces, no logos, no company names, no invented alphabets on signage. It should feel like one long exhale, not a slideshow. The last frame can carry one line of English text, my choice later — leave it clean. Give me the finished mp4, a cover frame, and the prompts you used per shot.

Inspect the task →

Website · 1 output

https://x.com/notdwd/status/2105373513104392552?s=20 reference this video for making production of https://evals-app.vercel.app/ site. launching demo video!

Inspect the task →

Website · 2 outputs

Create the Eiffel Tower and its surroundings in Paris as a detailed interactive 3D browser scene. Show recognizable lattice structure, all platforms, Champ de Mars paths and greenery, neighboring Parisian buildings, atmospheric lighting and orbit controls. Deliver a working browser artifact. Source: https://x.com/EnvolDev/status/2103535619603567054 (66K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 1 output

Build the highest-quality playable PUBG-style battle royale prototype you can in the browser. Include an original island map, movement, aiming, collectible equipment, simple computer opponents, a shrinking safe zone, health, win/lose and restart. Deliver a working game artifact with clear controls. Source: https://x.com/emmanuel_2m/status/2102717178223235228 (58K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 2 outputs

Create a detailed visual 3D presentation of an AH-64E Apache Guardian in Indian Air Force livery. Emphasize its recognizable silhouette, cockpit, rotor geometry, materials and dramatic lighting. Add orbit controls and rotor animation in a working browser artifact. This is a visual modeling exercise. Source: https://x.com/srikanthvaluri/status/2100935811059593547 (39K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 2 outputs

Create a 20-second concept launch video for ChatGPT-6 Sol and Luna. Use only these model names as factual copy; invent no performance, pricing or availability claims. Build a polished motion-design treatment with expressive typography, contrasting visual identities, elegant transitions and a final two-model lockup. Label it an unofficial concept. Deliver a playable animation and MP4 if available. Source: https://x.com/israelfemiojo/status/2102642002613481939 (196K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 2 outputs

Build a polished playable Sonic-style 3D platform game for the browser. Include a fast character, rings to collect, ramps, jumping, hazards, checkpoints, score and restart. Make speed and platforming feel satisfying. Deliver a working browser artifact. Adapt the source Godot task to the browser. Source: https://x.com/AiBattle_/status/2095994051354919049 (1.5M views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 1 output

Build a Night train. Create an immersive animated browser scene with a moving passenger train at night, lit windows, passing scenery, convincing wheels and tracks, and atmospheric light. Include a cinematic camera and pause/replay controls. Deliver a working browser artifact. Source: https://x.com/EnvolDev/status/2103282586055213355 (111K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 2 outputs

Build an interactive educational 3D jet engine in the browser. Include cutaway and x-ray views, animated airflow and rotating fan/compressor/turbine stages, clear labels, an exploded view and orbit controls. Explain the airflow visually and mark simplifications. Deliver the working browser artifact. Source: https://x.com/YouWareAI/status/2101271938706685991 (80K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 1 output

Build a detailed interactive 3D Roman Colosseum in the browser. Include the elliptical amphitheater, stacked arcades, recognizable ruined outer wall, arena floor and surrounding terrain. Add atmospheric lighting, orbit/zoom camera controls and an automatic cinematic tour. Deliver the working browser artifact. Source: https://x.com/LuminaBench/status/2104995514433564800 (65K views, 433 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 1 output

Create a bluetooth smart ring flex circuit using tscircuit. Deliver the tscircuit source, a rendered circuit/PCB preview and a concise component list. Explain any assumptions and validation limits; distinguish a concept design from manufacturing-ready hardware. Source: https://x.com/seveibar/status/2105043390576590978 (29.5K views, 153 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Create a playable 3D high-speed ride through a cyberpunk city in the browser. Include neon architecture, a convincing sense of speed, a controllable vehicle, obstacles, collision feedback, a score and restart. Support keyboard and visible touch controls, and deliver the working browser artifact. Source: https://x.com/higgsfield_ai/status/2105087030464245944 (37K views, 247 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 1 output

Build a beautiful interactive 3D voxel pagoda scene in the browser, with tiered curved roofs, red columns, detailed wooden structure, stone paths, trees and water in a coherent miniature landscape. Use voxel geometry, atmospheric lighting, and orbit/zoom controls. Deliver the working browser artifact. Source: https://x.com/LuminaBench/status/2104598576160456967 (40K views, 364 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Build a playable futuristic anti-gravity racing game inspired by the F-Zero-style racing demo in the source post. Create an original hovering vehicle, a twisting elevated track, boost, steering, lap timing and restart with a strong sense of speed. Deliver a browser-playable artifact. This adaptation targets the browser instead of the original Unreal environment. Source: https://x.com/ForwardEditor/status/2104633533981491575 (9.3K views, 128 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 1 output

Build a playful 3D pool party game with a colorful swimming pool, inflatable floats, animated water, and a controllable character that can move and jump into the pool with splash feedback. Add a simple collectible challenge, score and reset. Deliver a browser-playable artifact. This adaptation targets the browser instead of the original Blender/Godot workflow. Source: https://x.com/bijanbowen/status/2105113620891726041 (31.6K views, 354 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Create an interactive 3D gummy dice experience in the browser: colorful translucent dice with rounded edges, correctly arranged pips, glossy highlights, and soft stretchy jelly deformation as they drop, bounce and settle. Add a Roll button and appealing studio lighting. Deliver the working browser artifact. Source: https://x.com/vib3coded/status/2104828270789492913 (30K views, 559 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Create an interactive fall foliage simulator as a polished browser experience. Show a grove of trees whose leaves transition from green through gold, orange and red, then fall with believable wind and accumulate on the ground. Include season progression and wind controls, pause/reset, and a beautiful coherent composition. Deliver a working HTML artifact, not just an explanation. Source inspiration: https://x.com/claudeai/status/2104674987164782598 (Sonnet 5 vs Sonnet 5.5 fall foliage simulator; observed 687K views, 7.8K likes on 2026-10-01). This is an adapted reproduction brief based on the public post description, not the original verbatim prompt. The current comparison uses the models selected in Evals; do not claim to reproduce the original model versions.

Inspect the task →

Website · 2 outputs

Create a visually engaging animated history of dinosaurs in under one minute using JavaScript. Cover their emergence in the Triassic, the Jurassic giants, Cretaceous diversity, and the end-Cretaceous extinction, with clear English labels, a timeline, recognizable moving dinosaurs and smooth scene transitions. Deliver a playable HTML artifact with play/pause and replay controls. Source: https://x.com/claudeai/status/2104674993212702803 (273K views, 684 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Create a cinematic luxury car rental website with a striking premium hero, beautifully presented vehicle collection, subtle motion, refined typography, and a functional booking interaction with dates, vehicle choice and a clear price summary. Make the layout responsive and accessible. Use fictional branding and demo pricing. Deliver a working browser artifact. Source: https://x.com/viktoroddy/status/2105226806492283286 (25K views, 404 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Video · 2 outputs

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘

Inspect the task →

Video · 2 outputs

make a advertisment which can upload to instgram reels.

Inspect the task →

Video · 2 outputs

Make me a 15-second trailer for an F1 movie starring Ariana Grande

Inspect the task →

Video · 2 outputs

내가 지금 레퍼런스로 넣은 이미지들의 무드와 비슷하게 cozy한느낌의 브랜드 필름을 만들어줘

Inspect the task →

Video · 2 outputs

Please create a video showing the process of making a Rolex watch by an artisan,

Inspect the task →

Video · 2 outputs

make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out.

Inspect the task →

Video · 2 outputs

https://www.youtube.com/watch?v=o7DnBq7Fn74 The Powerpuff Girls x NewJeans music video starts at 1:20 in this video, and it combines the production process and graphics? Taking that feeling into account, Ain't In LA Song • ADÉLA, make a music video of the highlight part of the song for about 15 seconds

Inspect the task →

Video · 2 outputs

Make a 10-second cinematic promo video for Jeju Island: coastline, volcanic landscape and tangerine farms, ending on the text "Jeju. Slow down."

Inspect the task →

Website · 2 outputs

Build an interactive personal finance dashboard web app: monthly spending by category, a budget progress view, recent transactions, and a savings goal tracker, using realistic example data.

Inspect the task →

Video · 2 outputs

Make a 15-second animated announcement video for a coding bootcamp called "Launchpad" opening enrollment for its spring cohort, with the dates and a call to action.

Inspect the task →

Website · 2 outputs

Create a generative artwork inspired by a city at night seen from above: grids of streets, clusters of light, and traffic flow. Deliver it as an interactive piece that renders in the browser.

Inspect the task →

Video · 2 outputs

Make a 12-second animated video of a cute Norwegian Forest cat making a pretty strawberry drink at a café

Inspect the task →

Website · 2 outputs

Build the homepage for "Pixel Harbor", a three-person indie game studio. Show their upcoming game with a trailer placeholder, past titles, a devlog section, and a newsletter signup. It should feel like the studio has a strong personality.

Inspect the task →

Video · 2 outputs

Make a 15-second product teaser video for "Aura Buds", new wireless earbuds with 40-hour battery life and adaptive noise cancelling. End on the product name and "Available October 15".

Inspect the task →

Video · 2 outputs

Make a 30-second explainer video on why the sky is blue, for curious 12-year-olds, with clear visuals and on-screen captions.

Inspect the task →

Video · 2 outputs

Make a 20-second video slideshow of five surprising facts about octopuses, with a title card and a clear ending.

Inspect the task →

Website · 2 outputs

Create a generative art piece titled "Ocean Currents": flowing lines that suggest deep-sea currents, suitable for printing as a large wall poster. Deliver it as an interactive piece that renders in the browser.

Inspect the task →

Video · 2 outputs

Make a 20-second video that walks through a simple 4-step recipe for Korean egg fried rice, one step per slide, with a title and a final "Enjoy" card.

Inspect the task →

Website · 2 outputs

Build a pricing page for "Quill", an AI note-taking app for researchers. Three plans (Free, Pro, Lab), a monthly/yearly toggle, a feature comparison table, customer logos, and an FAQ.

Inspect the task →

Website · 2 outputs

Build a one-page landing site for "Ember & Oak", a small-batch specialty coffee roaster in Portland. Include a hero, the three signature roasts with tasting notes, a subscription offer, the roastery story, and a footer with location and hours.

Inspect the task →

Website · 2 outputs

Build a landing page for "Northbound", a travel company that runs small-group trips to Iceland, Norway and Greenland. Include a striking hero, three featured trips with imagery, testimonials, and a booking call to action.

Inspect the task →

Image · 2 outputs

Design the home screen of a mobile meditation app called "Still". Show today's session, a streak counter, mood check-in and recommended sessions. Deliver a high-fidelity mockup image of the screen.

Inspect the task →

Image · 2 outputs

Create a product photo for a bottle of cold-pressed orange juice from a brand called "Sunpress", styled for a summer social media ad.

Inspect the task →

Website · 2 outputs

Build a table reservation page for "Sora", a modern Korean bistro in Brooklyn: date, time and party-size selection, a short menu preview, opening hours, and a confirmation state after booking.

Inspect the task →

Website · 2 outputs

Build a portfolio website for Lena Park, an architect who designs small timber houses. Include a project grid, one project detail view, an about section, and a contact form.

Inspect the task →

Image · 2 outputs

Create a premium product photoshoot image for a handmade matte-black ceramic coffee mug, suitable for the hero of an online store.

Inspect the task →

Image · 2 outputs

Create a brand identity board for "Crumb & Co.", a neighborhood sourdough bakery: logo, color palette, typography, and a few example applications such as a bag, a storefront sign and an Instagram post. Deliver it as one image.

Inspect the task →

Loading answer…

Document · 2 outputs

Write a two-day walking guide to Kyoto for a first-time visitor, with a map-friendly list of stops.

Inspect the task →

Video · 2 outputs

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘 아이폰 광고 느낌 나게 말이야

Inspect the task →

Image · 2 outputs

Design a poster for a Friday jazz night at a small rooftop bar in Seoul.

Inspect the task →

Video · 2 outputs

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘 아이폰 광고 느낌 나게 말이야

Inspect the task →

Video · 1 output

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘 아이폰 광고 느낌 나게 말이야

Inspect the task →

Video · 1 output

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘 아이폰 광고 느낌 나게 말이야

Inspect the task →

Video · 2 outputs

I post office-humour memes and I want one with a hamster who has a job. 10 seconds, vertical 9:16. No dialogue — it has to be funny with the sound off. A round, soft hamster in a shirt slightly too big for it, at a desk built to its size: a monitor, a keyboard it has to walk across, a mug it holds with both paws. The joke is that it is taking the job completely seriously and the job is winning. Beat one: it types fast, very pleased with itself. Beat two: notifications pile up on the screen. Beat three: it stares straight down the lens, dead-eyed, cheeks full. Then it drinks from the mug like nothing happened. Keep it the same hamster the whole way through — same colour, same shirt, same desk. Lighting like a real office, not a cartoon: grey daylight, a monitor glow on the fur. Sound is keyboard clatter, a notification ping stack, one long sigh. Give me the finished mp4, a cover frame, and the prompts you used per shot.

Inspect the task →

App · 2 outputs

Make Super Mario as a 3D game I can actually play in the browser, and make playing video for 10seconds

Inspect the task →

Video · 2 outputs

I want a piece for the top of my feed that feels like a dream you half remember — dreamcore, but on the soft storybook side, not the horror side. 10 seconds, vertical 9:16, no people's faces, no text on screen. One continuous drift through three or four places that shouldn't connect but do: a bedroom where the wallpaper is sky, a corridor of doors lit by paper lanterns, a field of grass under two moons, a staircase going into shallow water. Everything slightly too still, everything slightly too bright at the edges. Colours washed and pastel — faded pink, pale green, milk blue — with the grain and soft halo of old film or a VHS tape. The camera never cuts. It floats forward, and the next place is already there when the last one fades. One small living thing moves in each place: a moth, a curtain, ripples, a paper boat. Sound is a muffled music-box tone and room tone, nothing sharp. Give me the finished mp4, a cover frame, and the prompts you used per shot.

Inspect the task →

Video · 1 output

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/higgsfield-ai-2103963588012744859.mp4

Inspect the task →

App · 2 outputs

Build a playable 2D samurai action game in the browser. It needs to pack a real punch. Basically, I want a trendy, cute anime girl as a 2D character, and I want to throw her into some intense battles. The animation should be flashy with a really kitschy vibe.

Inspect the task →

App · 2 outputs

Build an interactive cake-making game website where I can bake and decorate my own cake step by step. The UI should feel cute and playful, but the cake itself should look as realistic as possible, with lifelike textures for sponge, cream, and toppings.

Inspect the task →

Video · 2 outputs

프링글스 오리지널로 광고 만들어줘. 빨간 통에 콧수염 아저씨 그려진 그거. 통이 길쭉하고 칩이 안장처럼 휘어서 착착 쌓이는 게 이 제품의 전부야. 그 형태가 살아야 해. 사람이 먹는 장면은 빼줘. 제품만으로 15초를 끌고 가는 걸 보고 싶어. 뚜껑 여는 순간은 꼭 들어가야 해. 다들 그 소리랑 장면으로 기억하니까. 9:16 세로 15초. 소리 꺼도 이해되게.

Inspect the task →

Website · 2 outputs

이 영상 그대로 다시 만들어줘 → https://www.youtube.com/watch?v=ISYIEmQxs2M 비슷하게 말고 똑같이. 문구도 길이도 컷 넘어가는 지점도 그대로. 겉모양만 맞추지 말고 이즈 인 이즈 아웃 같은 모션 디테일까지 살려서. 등속으로 스윽 지나가면 다시 해야 해. 영상 파일 하나면 돼.

Inspect the task →

Video · 2 outputs

I run a small baking account and I want to try the tiny-food thing everyone keeps sending me — a whole strawberry tart made at doll size, start to finish. Vertical, 9:16, 15 seconds. One pair of hands, no face, no voice. Real miniature scale: a tart that ends up about the size of a coin, pastry rolled out with a pin the length of a matchstick, baked in a tiny fluted tin, and the tools should look like actual tools, not toys. Show the whole thing happening — pastry pressed into the tin, the tin going into the oven, the shell coming out, pastry cream piped in, then strawberries sliced and laid in a ring, and a glaze brushed over them at the end. The strawberries are the point, so they have to look like real fruit at that size: seeds on the skin, the pale core when they are cut, juice on the board. Sound is half the point: the knife through the fruit, the oven door, the brush on the glaze. Close and quiet, no music bed on top of it. End on the finished tart next to something everyday for scale — a coin, a fingertip, a matchstick — so the size lands.

Inspect the task →

Video · 2 outputs

Create a promotional video for Claude Opus 5.5. Use the official launch video in this post as a reference for the tone, pacing, and visual style: https://x.com/claudeai/status/2102435511222890900. Make it an original, polished piece with on-screen text that highlights what's new in Opus 5.5.

Inspect the task →

Video · 1 output

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/higgsfield-ai-2103963588012744859.mp4

Inspect the task →

Video · 2 outputs

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/achxvi-2104294892944097513.mp4

Inspect the task →

Video · 1 output

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/achxvi-2104137954427830588.mp4

Inspect the task →

Video · 1 output

make a demo video

Inspect the task →

Video · 2 outputs

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/shneural-2103151003272962130.mp4

Inspect the task →

App · 1 output

Build the best Minecraft clone you can in the browser: a playable voxel world with block breaking and placing, water physics, and realistic mechanics.

Inspect the task →

App · 2 outputs

Build a playable 2D samurai action game in the browser. It needs to pack a real punch. Basically, I want a trendy, cute anime girl as a 2D character, and I want to throw her into some intense battles. The animation should be flashy with a really kitschy vibe.

Inspect the task →

App · 2 outputs

Build an immersive 3D itinerary for an October 1–30 trip, with the links I'd need to plan it. Route: Chicago, Iceland, Amsterdam for World Summit AI, San Francisco for AI events, California national parks, Zion and Bryce Canyon, and a return through Las Vegas.

Inspect the task →

App · 2 outputs

Build a playable 3D samurai action game in the browser.

Inspect the task →

App · 1 output

Build the best Minecraft clone you can in the browser: a playable voxel world with block breaking and placing, water physics, and realistic mechanics.

Inspect the task →

App · 2 outputs

Build a polished browser game where the player escapes a miniature museum after closing: three connected rooms, an inventory, five linked puzzles, contextual hints, sound controls, and an ending.

Inspect the task →

Video · 2 outputs

make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out.

Inspect the task →

Video · 2 outputs

Create a motion graphic video. Give me the best. Make no mistakes.

Inspect the task →

Video · 2 outputs

Go pick a SaaS product and create a professional product launch video for it.

Inspect the task →

Website · 4 outputs

plz make dedicated website

Inspect the task →

Website · 2 outputs

Create a low-poly 3D island as a .glb model file, plus a set of 5 square PNG card-news images introducing the island.

Inspect the task →

Document · 2 outputs

- 저희가 "스킬샵" https://www.skills.ag/ 이라는 서비스를 새로 출시 했는데요, 디자인이 필요한 부분이 클로드 스킬을 판매하는 크레이터 분들에게 보낼 회사 소개서 / 제안서를 만들어야 합니다. 대략 8-10장 정도의 제안서이고, 클로드 스킬에 대해 익숙치 않으시다면 강의 / 워크숍 강사 구하는 느낌이라 생각하시면 됩니다 ! - 피그마에 대략적인 흐름을 화이트보딩 해놨습니다. 그 옆 오른쪽에는 클로드가 그냥 내뱉은거라 (카피라이팅이 좀 구리지만 레퍼용으로 해놓은) 그걸 기획서 보다는 그냥 쓸만한 문구나 아이디어 용으로 참고만 가볍게 하면 될것 같습니다. - 레퍼런스 폴더에 IT 기업 소개서 디자인 레퍼런스 몇개 보냅니다. 깔끔하고 미니멀한 느낌의 제안서 - 한눈에 보기 쉬운 느낌. 추가로 https://kmong.com/seller/become-a-seller, https://www.kmongcorp.com/ ((이러한 깔끔한 느낌), https://soomgo.team/ 이런 테크기업들도 참고하면 좋을거 같아요.

Inspect the task →

Image · 2 outputs

원피스를 한국에서 실사로 만든다면 누가 할지 캐스팅하고 포스터까지 만들어줘. 넷플릭스판은 서양 배우라 그 과장된 얼굴이 붙었는데, 한국 배우로 하면 어떨지 궁금해. 캐릭터 특징은 살리되 코스프레처럼 되면 안 돼. 루피 흉터랑 조로 눈 흉터 같은 건 그대로 가야 하고. 실제 한국 배우 중에서 골라서 누가 무슨 역인지 적어줘. 밀짚모자 일당 여섯 명 캐릭터 컷이랑 극장에 걸 메인 포스터 1장.

Inspect the task →

Website · 2 outputs

롯데 자이언츠 유니폼 새로 디자인해줘. 매년 조금씩만 바뀌어서 지금은 옛날 옷 같아. 요즘 나오는 유니폼처럼 가줘. 세로줄은 빼도 돼. 대신 뺐으면 그 자리를 뭘로 채웠는지 보여줘. 홈, 원정, 서드 3종과 착용 컷 2장. 등번호와 이름 들어갈 자리도 같이.

Inspect the task →

Website · 2 outputs

Build a Y2K-style interactive website where I can create glitter

Inspect the task →

Website · 2 outputs

Create a low-poly 3D island as a .glb model file, plus a set of 5

Inspect the task →

Benchmarks

Arena · WorkSep 9, 2026Study

Cost / quality frontier

13 models · score and cost

Claude Fable 5.1 (Max)14.73 score

Anthropiccost 6.93

Claude Opus 5 (High)11.27 score

Anthropiccost 3.91

Claude Fable 5 (High)9.58 score

Anthropiccost 3.34

GPT 5.6 Sol (xHigh)8.35 score

OpenAIcost 1.72

Claude Sonnet 5 (High)6.66 score

Anthropiccost 1.39

Kimi K3 (Max)6.66 score

Moonshotcost 1.06

GLM 5.2 (Max)5.54 score

Z.aicost 0.98

Hy4 preview5.49 score

Tencentcost 0.25

Deepseek V4 Flash (High) (20260731)2.84 score

DeepSeekcost 0.16

GLM 5.3 Flash1.8 score

Z.aicost 0.09

GPT 5.6 Luna (xHigh)1.06 score

OpenAIcost 0.07

Hy3-4.36 score

Tencentcost 0.04

Mimo V2.5 Pro-5.5 score

Xiaomicost 0.04

Arena · Work · published resultsRead original study ↗

SourceJun 8, 2026Study

On-policy distillation on GDPval

Distilled vs. base · 220 tasks per model

Distilled winsTiesBase wins

Nemotron 3 Nano+20.9 pp net win–loss

1232077

55.9% wins+5.1 weighted rubric pointsdistilled from Nemotron 3 Super

Qwen3.5-9B+21.4 pp net win–loss

1261579

57.3% wins+7.9 weighted rubric pointsdistilled from Qwen3.5-397B-A17B

Source · published resultsRead original study ↗

More comparisons

Website · 4 outputs

Build a playable 3D flying game in the browser (Three.js): pilot a paper plane through floating islands at golden hour. Soft volumetric clouds, waterfalls spilling off the islands, flocks of birds, glowing rings to fly through, wind trails, a cinematic intro fly-by and a score. Hand-painted storybook look, gorgeous from the first frame.

Inspect the task →

Website · 3 outputs

Build a playable 3D hover-bike racing game in the browser (Three.js). Neon cyberpunk city at night, rain, glowing billboards, bloom. Race through traffic with boost pads, ramps and near-miss sparks; speedometer, lap timer and a dramatic crash + replay screen. Make it look like a premium console game trailer the moment it loads.

Inspect the task →

Other · 4 outputs

Create a 12-second cinematic 3D product reveal film in the browser (Three.js) for a fictional smartwatch called ORBIT. Dark studio, dramatic rim lighting, slow camera orbit and dolly, liquid-metal reflections, a particle burst on the reveal, bold kinetic typography, logo end card. Auto-plays and loops like an Apple keynote launch film.

Inspect the task →

Website · 4 outputs

Build a playable 3D monster-truck smash game in the browser (Three.js + physics). Drive a giant monster truck through a colourful toy city and smash everything: towers of blocks topple, cars flip, glass shatters into debris with real physics, slow-motion on big crashes, dust and sparks, a destruction score. Should feel explosive and satisfying from the first second.

Inspect the task →

Loading answer…

Other · 3 outputs

Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.

Inspect the task →

Video · 2 outputs

https://www.instagram.com/reel/Dd4VIgTq8e7/ reference this video! you must reference the asset of the video. music, sound effect etc... and what u need to describe is https://evals.ag/ site!

Inspect the task →

Video · 2 outputs

Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.

Inspect the task →

Video · 1 output

Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.

Inspect the task →

Video · 1 output

make a demo video about https://evals.ag/

Inspect the task →

Video · 4 outputs

reference this video for motion graphic demo https://x.com/samgrows/status/2107571800020222393?s=20 you must describe this site https://evals.ag/

Inspect the task →

Video · 1 output

Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.

Inspect the task →

Video · 1 output

reference this video for motion graphic demo https://x.com/samgrows/status/2107571800020222393?s=20 you must describe this site https://evals.ag/

Inspect the task →

Video · 1 output

could you make a demo video reference this site https://x.com/jordanarchivess/status/2107536262894915762?s=20. what you need to make it site is https://evals.ag. you must use tts with elevenlabs v4 api

Inspect the task →

Video · 1 output

START NOW. Do not ask me any questions and do not wait for input — every input is below. Make every unspecified decision yourself, log it in DECISIONS.md, and keep going until ONE finished video file exists. The only deliverable is that one .mp4. Task Make ONE ~60s landscape demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. Output: /Users/evan/workspace/video-generator/videos/evals-demo3/renders/evals-demo3.mp4, 1920×1080, 30fps, H.264 + AAC 192k, −14 LUFS. Nothing else to deliver (no vertical version, no variants). Work dir: repo /Users/evan/workspace/video-generator (Claude Code on this Mac). Build in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-demo3/. Render: npx --yes hyperframes@0.8.101 render -o <out> --quiet from the project dir. Use curl for HTTP/API calls (python has no SSL certs). ElevenLabs key: ELEVENLABS_API_KEY in /Users/evan/workspace/video-generator/.env. Story (English voiceover + on-screen type) Hook, first 3s must be instantly clear: real AI-made GAME/VIDEO outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." Type a real game/video task into the Evals composer → Battle → up to 4 setups (Model · Skill · Environment, each landing with real values and moving UI). Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. Blind compare: two moving outputs, "Which would you choose?" → vote ("I prefer A"). Run record of that same output (real LLM, skill, environment, turn time, tokens). Setup leaderboard ("Setups, not just models."). Community wall of real game/video/shorts outputs. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. Content rules Show ONLY videos, games, 3D scenes, motion videos, shorts made in Evals. No websites / landing pages. Main thread = the real 4-setup Sonic battle: prompt "Build a polished playable Sonic-style 3D platform game for the browser.", clips /Users/evan/workspace/video-generator/videos/evals-launch/assets/clips/run_a.mp4 … run_d.mp4, run https://evals-app.vercel.app/runs/run_597Q2CRE4FXNJ2R7HZXMVF4Z03. Capture fresh real UI from https://evals-app.vercel.app with Playwright at 2× DPR: composer, Run-mode menu, setups stepper, compare page, output run record, Leaderboards, Explore → Outputs. If a page needs sign-in, use the public pages. Record extra community game/video outputs from their public "Open preview" links. Screencast 1600×900, 30fps, 6–7s, with keys pressed or orbit so they move. Reuse the pattern in /Users/evan/workspace/video-generator/videos/evals-launch/scripts/rec.mjs. No real-person likeness, no invented features/numbers/names. Only show model names you see in captures (e.g. Claude Opus 5.5, GPT-6 Astra). Voiceover (must flow continuously) ElevenLabs voice Will bIHbv24MWmeRgasZH58o, model eleven_v4, stability 0.15, similarity 0.8, style 0.8, speed 1.1, with emotion tags in the text ([excited] [confident] [curious] [emphatic] [quick]). Use the /with-timestamps endpoint. Write ~165–180 words so the voice fills the whole film. Generate it as ONE take and lay it unspliced — never cut it into phrases or time-stretch pieces. Longest silence inside the VO ≤ 0.35s. No clipped words. The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. Verify with Whisper that every word is there and no tags are spoken. If "Evals" sounds like "evils", respell it "E-vals". Music & sound New modern premium BGM, not cheesy. Generate 2–3 candidates with ElevenLabs Music: "modern minimal tech launch, punchy kick, deep 808, crisp hats, glassy synth plucks, Apple/Linear keynote feel, no cheesy EDM risers, no corporate/stock vibe, 120 BPM". Pick the cleanest; detect BPM and the first downbeat. VO leads at about −16 LUFS. Music ducks about 6dB only while he speaks (smooth envelope). SFX bus ~3dB under the music, above it only on hits. At 4–6 big hits, layer reverse swell + boom (HP 60Hz) + transient + glitch. Light ticks and whooshes elsewhere — never on every cut. Look Fresh premium design system, written once in DESIGN.md: one sans font, one accent colour real UI captures as rounded cards (16–22px) with soft shadows dark/light section switches Never: outlines/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups. Motion (non-negotiable — this is the feel) Reference for feel: /Users/evan/workspace/video-generator/videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4; rules in /Users/evan/workspace/video-generator/videos/evals-hype/MOTION_FIX.md. One camera = a pure function of time with velocity-continuous easing. Deterministic: no timers or randomness. Two easing families only: snappy ≈25% of remaining distance per frame, floaty ≈12%. Exits accelerate into the cut; the next shot carries the velocity through it. Real whip-pans: both shots travel together over 0.3–0.45s with velocity blur. Overlays are screen-locked. Overlapping action, no start-stop chains. Every hold drifts 1–3%. No dead stops, no frozen frames, no double cuts. Word-by-word reveals 2–4 frames apart; the newest word flashes the accent. Cuts on the beat ±1 frame. Speed by re-rendering: render at --fps 30/speedup (e.g. --fps 95/4 = 1.263×), play back at 30fps. Impact FX in post (ffmpeg) on 4–6 big hits only: +8.5% zoom punch (τ 0.09s) 16px shake at ~22Hz (τ 0.1s) rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash A lighter version on secondary cuts. QA (do it, then deliver) python3 /Users/evan/workspace/video-generator/videos/evals-hype/scripts/motion.py <render> <bpm> <first_beat>: no dead stops, no frozen holds, cuts on the beat. sh /Users/evan/workspace/video-generator/videos/evals-hype/scripts/strip.sh <render> <t> <out.png> at every transition — look at each strip. VO RMS gap check (≤ 0.35s) plus a Whisper pass on the final mix. Text readable and inside the margins; every name and number on screen is real. Fix anything that fails, re-render, and finish with the single file videos/evals-demo3/renders/evals-demo3.mp4.

Inspect the task →

Other · 2 outputs

Make a ~60s product launch film for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. 16:9, 1920×1080, 30fps, music-locked, no narration — on-screen type tells the story. Build it in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-<name>/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use my logged-in Chrome session for anything behind sign-in). Motion reference: videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its feel exactly; change only scenario, design and assets. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Download the reference and study it frame by frame. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. strip.sh (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver renders/evals-<name>-v1.mp4 plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.

Inspect the task →

Video · 2 outputs

Make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464 You must reference this site https://evals.ag/. You must use font sanserif. Motion must be speedy. Never make ai slop.

Inspect the task →

Video · 2 outputs

Make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464 You must reference this site https://evals.ag/. You must use font sanserif. Motion must be speedy. Never make ai slop.

Inspect the task →

Video · 1 output

Evals landscape demo-video prompt (approved "감도") Copy everything below the line into a new session. Make a ~60s landscape product demo video for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. Horizontal 16:9, 1920×1080, 30fps (for YouTube / website / X timeline). Not a vertical Short/Reel — no 9:16 output, no vertical crops, no phone safe-zones. Music-locked, no narration — on-screen type tells the story while the real product is demoed end to end. Build it in HyperFrames (HTML + one paused GSAP timeline) under /Users/evan/workspace/video-generator/videos/evals-demo3/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use the logged-in Claude in Chrome session if connected; otherwise public pages). Motion reference (local file, already downloaded): /Users/evan/workspace/video-generator/videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its motion feel exactly (it is also 16:9); change only scenario, design and assets. Its 26s cut-down is just the reference for feel — this video runs the full ~60s. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." Run it autonomously — do not ask me for input Work in Claude Code with the repo /Users/evan/workspace/video-generator as the working directory. Every path below is absolute; everything you need is already on this machine. Do NOT stop to ask questions or request inputs. Every input is specified here; for anything not specified, pick a sensible default, write the decision into DECISIONS.md, and keep going until the final file is rendered. If something is unavailable, use the fallback and continue: Chrome not connected or a page needs sign-in → use the public pages (https://evals-app.vercel.app home, /compare, leaderboards, explore) with Playwright, and record outputs from their public "Open preview" links. Can't run a new Battle → use public outputs that share the same prompt (the /compare pages show pairs). Music: pick an unused track from /Users/evan/workspace/video-generator/videos/evals-hype/music3/ or music2/. SFX: /Users/evan/workspace/video-generator/videos/evals-hype/sfx2/, ref2/sfx3/, ref2/sfx4/. QA scripts: /Users/evan/workspace/video-generator/videos/evals-hype/scripts/motion.py and strip.sh (also beats.py for BPM). If they fail, write equivalents. Render: npx --yes hyperframes@0.8.101 render -o <out.mp4> --quiet from the project dir (--fps 95/4 etc. for speed re-render). API calls via curl (python has no SSL certs here). What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Study the local reference file frame by frame (ffmpeg frame dumps + frame-diff). Its design notes are in /Users/evan/workspace/video-generator/videos/evals-hype/ref2/SPEC.md and MOTION_FIX.md — read them, but design and copy must be new. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) python3 /Users/evan/workspace/video-generator/videos/evals-hype/scripts/motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. sh /Users/evan/workspace/video-generator/videos/evals-hype/scripts/strip.sh <render> <t> <out.png> (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver one 1920×1080 file renders/evals-demo3-v1.mp4 (no vertical version) plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.

Inspect the task →

Make a ~60s product launch film for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. 16:9, 1920×1080, 30fps, music-locked, no narration — on-screen type tells the story. Build it in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-<name>/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use my logged-in Chrome session for anything behind sign-in). Motion reference: videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its feel exactly; change only scenario, design and assets. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Download the reference and study it frame by frame. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. strip.sh (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver renders/evals-<name>-v1.mp4 plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.

Inspect the task →

Video · 2 outputs

Make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464 You must reference this site https://evals.ag/. You must use font sanserif. Motion must be speedy. Never make ai slop.

Inspect the task →

Video · 2 outputs

Make me a 15-second trailer for an F1 movie starring Ariana Grande

Inspect the task →

Video · 2 outputs

내가 지금 레퍼런스로 넣은 이미지들의 무드와 비슷하게 cozy한느낌의 브랜드 필름을 만들어줘

Inspect the task →

Video · 1 output

make a viral video in tiktok trending!

Inspect the task →

Video · 2 outputs

make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464?s=20 what you make it is https://evals.ag/ site

Inspect the task →

Website · 2 outputs

Build a single-page landing site for a coffee subscription startup with a hero, pricing table and signup form.

Inspect the task →

Video · 2 outputs

make a demo intro video reference this video https://x.com/rajsinghfirst/status/2107161801322549725?s=20 demo site is https://evals.ag ! this site

Inspect the task →

Video · 2 outputs

make a viral video reference this thing https://x.com/motion_conquest/status/2106786068712558848?s=20 what you need to make it is https://evals.ag/ site! demo vide

Inspect the task →

Video · 2 outputs

make a demo intro video reference this video https://x.com/rajsinghfirst/status/2107161801322549725?s=20 demo site is https://evals.ag ! this site

Inspect the task →

Other · 2 outputs

make a ai influencer for x viral

Inspect the task →

Video · 2 outputs

make a viral video in x

Inspect the task →

Website · 2 outputs

Build a landing page for a neighborhood bakery: menu, opening hours, map and a reservation button. It should look great on phones.

Inspect the task →

Website · 1 output

make a 2d kawai samurai game.

Inspect the task →

Website · 4 outputs

Build a polished playable Sonic-style 3D platform game for the browser. Include a fast character, rings to collect, ramps, jumping, hazards, checkpoints, score and restart. Make speed and platforming feel satisfying. Deliver a working browser artifact.

Inspect the task →

Website · 2 outputs

Build a polished playable Sonic-style 3D platform game for the browser. Include a fast character, rings to collect, ramps, jumping, hazards, checkpoints, score and restart. Make speed and platforming feel satisfying. Deliver a working browser artifact.

Inspect the task →

Document · 2 outputs

1+1?

Inspect the task →

Video · 2 outputs

Make me a 15-second ad with MapleStory characters built in 3D

Inspect the task →

Website · 2 outputs

Make MapleStory as a 3D game I can play in the browser

Inspect the task →

Video · 2 outputs

I want the whole human story in one shot — from the first people to the AI we use now. 15 seconds, 16:9, no narration, no on-screen text until the very last frame. One continuous move, never cutting. The camera pushes forward and the world changes around it: firelight on a cave wall, a hand pressed in ochre, then fields and the first walls, then a printing press, then steel and smoke, then a wall of city light, then a room of humming machines, then the light of a screen on a face in the dark. Each era arrives while the last one is still leaving — the fire becomes a furnace, the cave painting becomes a page, the page becomes a screen. Hand things off on shape: something round stays round, something bright stays in the same corner of the frame. Keep a human in almost every era, always small in the frame, so the scale of the thing reads against a person. No famous faces, no logos, no company names, no invented alphabets on signage. It should feel like one long exhale, not a slideshow. The last frame can carry one line of English text, my choice later — leave it clean. Give me the finished mp4, a cover frame, and the prompts you used per shot.

Inspect the task →

Website · 1 output

https://x.com/notdwd/status/2105373513104392552?s=20 reference this video for making production of https://evals-app.vercel.app/ site. launching demo video!

Inspect the task →

Website · 2 outputs

Create the Eiffel Tower and its surroundings in Paris as a detailed interactive 3D browser scene. Show recognizable lattice structure, all platforms, Champ de Mars paths and greenery, neighboring Parisian buildings, atmospheric lighting and orbit controls. Deliver a working browser artifact. Source: https://x.com/EnvolDev/status/2103535619603567054 (66K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 1 output

Build the highest-quality playable PUBG-style battle royale prototype you can in the browser. Include an original island map, movement, aiming, collectible equipment, simple computer opponents, a shrinking safe zone, health, win/lose and restart. Deliver a working game artifact with clear controls. Source: https://x.com/emmanuel_2m/status/2102717178223235228 (58K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 2 outputs

Create a detailed visual 3D presentation of an AH-64E Apache Guardian in Indian Air Force livery. Emphasize its recognizable silhouette, cockpit, rotor geometry, materials and dramatic lighting. Add orbit controls and rotor animation in a working browser artifact. This is a visual modeling exercise. Source: https://x.com/srikanthvaluri/status/2100935811059593547 (39K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 2 outputs

Create a 20-second concept launch video for ChatGPT-6 Sol and Luna. Use only these model names as factual copy; invent no performance, pricing or availability claims. Build a polished motion-design treatment with expressive typography, contrasting visual identities, elegant transitions and a final two-model lockup. Label it an unofficial concept. Deliver a playable animation and MP4 if available. Source: https://x.com/israelfemiojo/status/2102642002613481939 (196K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 2 outputs

Build a polished playable Sonic-style 3D platform game for the browser. Include a fast character, rings to collect, ramps, jumping, hazards, checkpoints, score and restart. Make speed and platforming feel satisfying. Deliver a working browser artifact. Adapt the source Godot task to the browser. Source: https://x.com/AiBattle_/status/2095994051354919049 (1.5M views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 1 output

Build a Night train. Create an immersive animated browser scene with a moving passenger train at night, lit windows, passing scenery, convincing wheels and tracks, and atmospheric light. Include a cinematic camera and pause/replay controls. Deliver a working browser artifact. Source: https://x.com/EnvolDev/status/2103282586055213355 (111K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 2 outputs

Build an interactive educational 3D jet engine in the browser. Include cutaway and x-ray views, animated airflow and rotating fan/compressor/turbine stages, clear labels, an exploded view and orbit controls. Explain the airflow visually and mark simplifications. Deliver the working browser artifact. Source: https://x.com/YouWareAI/status/2101271938706685991 (80K views observed 2026-10-01). Adapted recreation brief based on the public post, not a verbatim original prompt or replication of its model versions. Use the models selected in Evals.

Inspect the task →

Website · 1 output

Build a detailed interactive 3D Roman Colosseum in the browser. Include the elliptical amphitheater, stacked arcades, recognizable ruined outer wall, arena floor and surrounding terrain. Add atmospheric lighting, orbit/zoom camera controls and an automatic cinematic tour. Deliver the working browser artifact. Source: https://x.com/LuminaBench/status/2104995514433564800 (65K views, 433 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 1 output

Create a bluetooth smart ring flex circuit using tscircuit. Deliver the tscircuit source, a rendered circuit/PCB preview and a concise component list. Explain any assumptions and validation limits; distinguish a concept design from manufacturing-ready hardware. Source: https://x.com/seveibar/status/2105043390576590978 (29.5K views, 153 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Create a playable 3D high-speed ride through a cyberpunk city in the browser. Include neon architecture, a convincing sense of speed, a controllable vehicle, obstacles, collision feedback, a score and restart. Support keyboard and visible touch controls, and deliver the working browser artifact. Source: https://x.com/higgsfield_ai/status/2105087030464245944 (37K views, 247 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 1 output

Build a beautiful interactive 3D voxel pagoda scene in the browser, with tiered curved roofs, red columns, detailed wooden structure, stone paths, trees and water in a coherent miniature landscape. Use voxel geometry, atmospheric lighting, and orbit/zoom controls. Deliver the working browser artifact. Source: https://x.com/LuminaBench/status/2104598576160456967 (40K views, 364 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Build a playable futuristic anti-gravity racing game inspired by the F-Zero-style racing demo in the source post. Create an original hovering vehicle, a twisting elevated track, boost, steering, lap timing and restart with a strong sense of speed. Deliver a browser-playable artifact. This adaptation targets the browser instead of the original Unreal environment. Source: https://x.com/ForwardEditor/status/2104633533981491575 (9.3K views, 128 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 1 output

Build a playful 3D pool party game with a colorful swimming pool, inflatable floats, animated water, and a controllable character that can move and jump into the pool with splash feedback. Add a simple collectible challenge, score and reset. Deliver a browser-playable artifact. This adaptation targets the browser instead of the original Blender/Godot workflow. Source: https://x.com/bijanbowen/status/2105113620891726041 (31.6K views, 354 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Create an interactive 3D gummy dice experience in the browser: colorful translucent dice with rounded edges, correctly arranged pips, glossy highlights, and soft stretchy jelly deformation as they drop, bounce and settle. Add a Roll button and appealing studio lighting. Deliver the working browser artifact. Source: https://x.com/vib3coded/status/2104828270789492913 (30K views, 559 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Create an interactive fall foliage simulator as a polished browser experience. Show a grove of trees whose leaves transition from green through gold, orange and red, then fall with believable wind and accumulate on the ground. Include season progression and wind controls, pause/reset, and a beautiful coherent composition. Deliver a working HTML artifact, not just an explanation. Source inspiration: https://x.com/claudeai/status/2104674987164782598 (Sonnet 5 vs Sonnet 5.5 fall foliage simulator; observed 687K views, 7.8K likes on 2026-10-01). This is an adapted reproduction brief based on the public post description, not the original verbatim prompt. The current comparison uses the models selected in Evals; do not claim to reproduce the original model versions.

Inspect the task →

Website · 2 outputs

Create a visually engaging animated history of dinosaurs in under one minute using JavaScript. Cover their emergence in the Triassic, the Jurassic giants, Cretaceous diversity, and the end-Cretaceous extinction, with clear English labels, a timeline, recognizable moving dinosaurs and smooth scene transitions. Deliver a playable HTML artifact with play/pause and replay controls. Source: https://x.com/claudeai/status/2104674993212702803 (273K views, 684 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Website · 2 outputs

Create a cinematic luxury car rental website with a striking premium hero, beautifully presented vehicle collection, subtle motion, refined typography, and a functional booking interaction with dates, vehicle choice and a clear price summary. Make the layout responsive and accessible. Use fictional branding and demo pricing. Deliver a working browser artifact. Source: https://x.com/viktoroddy/status/2105226806492283286 (25K views, 404 likes observed 2026-10-01). Adapted reproduction brief based on the public post; not the original verbatim prompt. Compare the current Evals-selected models, not the original model versions.

Inspect the task →

Video · 2 outputs

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘

Inspect the task →

Video · 2 outputs

make a advertisment which can upload to instgram reels.

Inspect the task →

Video · 2 outputs

Make me a 15-second trailer for an F1 movie starring Ariana Grande

Inspect the task →

Video · 2 outputs

내가 지금 레퍼런스로 넣은 이미지들의 무드와 비슷하게 cozy한느낌의 브랜드 필름을 만들어줘

Inspect the task →

Video · 2 outputs

Please create a video showing the process of making a Rolex watch by an artisan,

Inspect the task →

Video · 2 outputs

make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out.

Inspect the task →

Video · 2 outputs

https://www.youtube.com/watch?v=o7DnBq7Fn74 The Powerpuff Girls x NewJeans music video starts at 1:20 in this video, and it combines the production process and graphics? Taking that feeling into account, Ain't In LA Song • ADÉLA, make a music video of the highlight part of the song for about 15 seconds

Inspect the task →

Video · 2 outputs

Make a 10-second cinematic promo video for Jeju Island: coastline, volcanic landscape and tangerine farms, ending on the text "Jeju. Slow down."

Inspect the task →

Website · 2 outputs

Build an interactive personal finance dashboard web app: monthly spending by category, a budget progress view, recent transactions, and a savings goal tracker, using realistic example data.

Inspect the task →

Video · 2 outputs

Make a 15-second animated announcement video for a coding bootcamp called "Launchpad" opening enrollment for its spring cohort, with the dates and a call to action.

Inspect the task →

Website · 2 outputs

Create a generative artwork inspired by a city at night seen from above: grids of streets, clusters of light, and traffic flow. Deliver it as an interactive piece that renders in the browser.

Inspect the task →

Video · 2 outputs

Make a 12-second animated video of a cute Norwegian Forest cat making a pretty strawberry drink at a café

Inspect the task →

Website · 2 outputs

Build the homepage for "Pixel Harbor", a three-person indie game studio. Show their upcoming game with a trailer placeholder, past titles, a devlog section, and a newsletter signup. It should feel like the studio has a strong personality.

Inspect the task →

Video · 2 outputs

Make a 15-second product teaser video for "Aura Buds", new wireless earbuds with 40-hour battery life and adaptive noise cancelling. End on the product name and "Available October 15".

Inspect the task →

Video · 2 outputs

Make a 30-second explainer video on why the sky is blue, for curious 12-year-olds, with clear visuals and on-screen captions.

Inspect the task →

Video · 2 outputs

Make a 20-second video slideshow of five surprising facts about octopuses, with a title card and a clear ending.

Inspect the task →

Website · 2 outputs

Create a generative art piece titled "Ocean Currents": flowing lines that suggest deep-sea currents, suitable for printing as a large wall poster. Deliver it as an interactive piece that renders in the browser.

Inspect the task →

Video · 2 outputs

Make a 20-second video that walks through a simple 4-step recipe for Korean egg fried rice, one step per slide, with a title and a final "Enjoy" card.

Inspect the task →

Website · 2 outputs

Build a pricing page for "Quill", an AI note-taking app for researchers. Three plans (Free, Pro, Lab), a monthly/yearly toggle, a feature comparison table, customer logos, and an FAQ.

Inspect the task →

Website · 2 outputs

Build a one-page landing site for "Ember & Oak", a small-batch specialty coffee roaster in Portland. Include a hero, the three signature roasts with tasting notes, a subscription offer, the roastery story, and a footer with location and hours.

Inspect the task →

Website · 2 outputs

Build a landing page for "Northbound", a travel company that runs small-group trips to Iceland, Norway and Greenland. Include a striking hero, three featured trips with imagery, testimonials, and a booking call to action.

Inspect the task →

Image · 2 outputs

Design the home screen of a mobile meditation app called "Still". Show today's session, a streak counter, mood check-in and recommended sessions. Deliver a high-fidelity mockup image of the screen.

Inspect the task →

Image · 2 outputs

Create a product photo for a bottle of cold-pressed orange juice from a brand called "Sunpress", styled for a summer social media ad.

Inspect the task →

Website · 2 outputs

Build a table reservation page for "Sora", a modern Korean bistro in Brooklyn: date, time and party-size selection, a short menu preview, opening hours, and a confirmation state after booking.

Inspect the task →

Website · 2 outputs

Build a portfolio website for Lena Park, an architect who designs small timber houses. Include a project grid, one project detail view, an about section, and a contact form.

Inspect the task →

Image · 2 outputs

Create a premium product photoshoot image for a handmade matte-black ceramic coffee mug, suitable for the hero of an online store.

Inspect the task →

Image · 2 outputs

Create a brand identity board for "Crumb & Co.", a neighborhood sourdough bakery: logo, color palette, typography, and a few example applications such as a bag, a storefront sign and an Instagram post. Deliver it as one image.

Inspect the task →

Loading answer…

Document · 2 outputs

Write a two-day walking guide to Kyoto for a first-time visitor, with a map-friendly list of stops.

Inspect the task →

Video · 2 outputs

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘 아이폰 광고 느낌 나게 말이야

Inspect the task →

Image · 2 outputs

Design a poster for a Friday jazz night at a small rooftop bar in Seoul.

Inspect the task →

Video · 2 outputs

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘 아이폰 광고 느낌 나게 말이야

Inspect the task →

Video · 1 output

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘 아이폰 광고 느낌 나게 말이야

Inspect the task →

Video · 1 output

https://www.skills.ag/skills/skill_01KRJP787GB1PGW20M61REAJ69 이 사이트 보고 제품 설명 줬잖아 인스타 릴스 광고 만들어줘 아이폰 광고 느낌 나게 말이야

Inspect the task →

Video · 2 outputs

I post office-humour memes and I want one with a hamster who has a job. 10 seconds, vertical 9:16. No dialogue — it has to be funny with the sound off. A round, soft hamster in a shirt slightly too big for it, at a desk built to its size: a monitor, a keyboard it has to walk across, a mug it holds with both paws. The joke is that it is taking the job completely seriously and the job is winning. Beat one: it types fast, very pleased with itself. Beat two: notifications pile up on the screen. Beat three: it stares straight down the lens, dead-eyed, cheeks full. Then it drinks from the mug like nothing happened. Keep it the same hamster the whole way through — same colour, same shirt, same desk. Lighting like a real office, not a cartoon: grey daylight, a monitor glow on the fur. Sound is keyboard clatter, a notification ping stack, one long sigh. Give me the finished mp4, a cover frame, and the prompts you used per shot.

Inspect the task →

App · 2 outputs

Make Super Mario as a 3D game I can actually play in the browser, and make playing video for 10seconds

Inspect the task →

Video · 2 outputs

I want a piece for the top of my feed that feels like a dream you half remember — dreamcore, but on the soft storybook side, not the horror side. 10 seconds, vertical 9:16, no people's faces, no text on screen. One continuous drift through three or four places that shouldn't connect but do: a bedroom where the wallpaper is sky, a corridor of doors lit by paper lanterns, a field of grass under two moons, a staircase going into shallow water. Everything slightly too still, everything slightly too bright at the edges. Colours washed and pastel — faded pink, pale green, milk blue — with the grain and soft halo of old film or a VHS tape. The camera never cuts. It floats forward, and the next place is already there when the last one fades. One small living thing moves in each place: a moth, a curtain, ripples, a paper boat. Sound is a muffled music-box tone and room tone, nothing sharp. Give me the finished mp4, a cover frame, and the prompts you used per shot.

Inspect the task →

Video · 1 output

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/higgsfield-ai-2103963588012744859.mp4

Inspect the task →

App · 2 outputs

Build a playable 2D samurai action game in the browser. It needs to pack a real punch. Basically, I want a trendy, cute anime girl as a 2D character, and I want to throw her into some intense battles. The animation should be flashy with a really kitschy vibe.

Inspect the task →

App · 2 outputs

Build an interactive cake-making game website where I can bake and decorate my own cake step by step. The UI should feel cute and playful, but the cake itself should look as realistic as possible, with lifelike textures for sponge, cream, and toppings.

Inspect the task →

Video · 2 outputs

프링글스 오리지널로 광고 만들어줘. 빨간 통에 콧수염 아저씨 그려진 그거. 통이 길쭉하고 칩이 안장처럼 휘어서 착착 쌓이는 게 이 제품의 전부야. 그 형태가 살아야 해. 사람이 먹는 장면은 빼줘. 제품만으로 15초를 끌고 가는 걸 보고 싶어. 뚜껑 여는 순간은 꼭 들어가야 해. 다들 그 소리랑 장면으로 기억하니까. 9:16 세로 15초. 소리 꺼도 이해되게.

Inspect the task →

Website · 2 outputs

이 영상 그대로 다시 만들어줘 → https://www.youtube.com/watch?v=ISYIEmQxs2M 비슷하게 말고 똑같이. 문구도 길이도 컷 넘어가는 지점도 그대로. 겉모양만 맞추지 말고 이즈 인 이즈 아웃 같은 모션 디테일까지 살려서. 등속으로 스윽 지나가면 다시 해야 해. 영상 파일 하나면 돼.

Inspect the task →

Video · 2 outputs

I run a small baking account and I want to try the tiny-food thing everyone keeps sending me — a whole strawberry tart made at doll size, start to finish. Vertical, 9:16, 15 seconds. One pair of hands, no face, no voice. Real miniature scale: a tart that ends up about the size of a coin, pastry rolled out with a pin the length of a matchstick, baked in a tiny fluted tin, and the tools should look like actual tools, not toys. Show the whole thing happening — pastry pressed into the tin, the tin going into the oven, the shell coming out, pastry cream piped in, then strawberries sliced and laid in a ring, and a glaze brushed over them at the end. The strawberries are the point, so they have to look like real fruit at that size: seeds on the skin, the pale core when they are cut, juice on the board. Sound is half the point: the knife through the fruit, the oven door, the brush on the glaze. Close and quiet, no music bed on top of it. End on the finished tart next to something everyday for scale — a coin, a fingertip, a matchstick — so the size lands.

Inspect the task →

Video · 2 outputs

Create a promotional video for Claude Opus 5.5. Use the official launch video in this post as a reference for the tone, pacing, and visual style: https://x.com/claudeai/status/2102435511222890900. Make it an original, polished piece with on-screen text that highlights what's new in Opus 5.5.

Inspect the task →

Video · 1 output

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/higgsfield-ai-2103963588012744859.mp4

Inspect the task →

Video · 2 outputs

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/achxvi-2104294892944097513.mp4

Inspect the task →

Video · 1 output

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/achxvi-2104137954427830588.mp4

Inspect the task →

Video · 1 output

make a demo video

Inspect the task →

Video · 2 outputs

Repurpose this source video into a fresh short-form video. Keep the core idea and pacing, but make the visuals, script, and structure original. Source video: https://cdn.revid.ai/static/claude-motion-graphics/shneural-2103151003272962130.mp4

Inspect the task →

App · 1 output

Build the best Minecraft clone you can in the browser: a playable voxel world with block breaking and placing, water physics, and realistic mechanics.

Inspect the task →

App · 2 outputs

Build a playable 2D samurai action game in the browser. It needs to pack a real punch. Basically, I want a trendy, cute anime girl as a 2D character, and I want to throw her into some intense battles. The animation should be flashy with a really kitschy vibe.

Inspect the task →

App · 2 outputs

Build an immersive 3D itinerary for an October 1–30 trip, with the links I'd need to plan it. Route: Chicago, Iceland, Amsterdam for World Summit AI, San Francisco for AI events, California national parks, Zion and Bryce Canyon, and a return through Las Vegas.

Inspect the task →

App · 2 outputs

Build a playable 3D samurai action game in the browser.

Inspect the task →

App · 1 output

Build the best Minecraft clone you can in the browser: a playable voxel world with block breaking and placing, water physics, and realistic mechanics.

Inspect the task →

App · 2 outputs

Build a polished browser game where the player escapes a miniature museum after closing: three connected rooms, an inventory, five linked puzzles, contextual hints, sound controls, and an ending.

Inspect the task →

Video · 2 outputs

make a dynamic 15-second motion graphics video that shows what an incredible motion designer you are, like it's your showreel for a résumé. go all out.

Inspect the task →

Video · 2 outputs

Create a motion graphic video. Give me the best. Make no mistakes.

Inspect the task →

Video · 2 outputs

Go pick a SaaS product and create a professional product launch video for it.

Inspect the task →

Website · 4 outputs

plz make dedicated website

Inspect the task →

Website · 2 outputs

Create a low-poly 3D island as a .glb model file, plus a set of 5 square PNG card-news images introducing the island.

Inspect the task →

Document · 2 outputs

- 저희가 "스킬샵" https://www.skills.ag/ 이라는 서비스를 새로 출시 했는데요, 디자인이 필요한 부분이 클로드 스킬을 판매하는 크레이터 분들에게 보낼 회사 소개서 / 제안서를 만들어야 합니다. 대략 8-10장 정도의 제안서이고, 클로드 스킬에 대해 익숙치 않으시다면 강의 / 워크숍 강사 구하는 느낌이라 생각하시면 됩니다 ! - 피그마에 대략적인 흐름을 화이트보딩 해놨습니다. 그 옆 오른쪽에는 클로드가 그냥 내뱉은거라 (카피라이팅이 좀 구리지만 레퍼용으로 해놓은) 그걸 기획서 보다는 그냥 쓸만한 문구나 아이디어 용으로 참고만 가볍게 하면 될것 같습니다. - 레퍼런스 폴더에 IT 기업 소개서 디자인 레퍼런스 몇개 보냅니다. 깔끔하고 미니멀한 느낌의 제안서 - 한눈에 보기 쉬운 느낌. 추가로 https://kmong.com/seller/become-a-seller, https://www.kmongcorp.com/ ((이러한 깔끔한 느낌), https://soomgo.team/ 이런 테크기업들도 참고하면 좋을거 같아요.

Inspect the task →

Image · 2 outputs

원피스를 한국에서 실사로 만든다면 누가 할지 캐스팅하고 포스터까지 만들어줘. 넷플릭스판은 서양 배우라 그 과장된 얼굴이 붙었는데, 한국 배우로 하면 어떨지 궁금해. 캐릭터 특징은 살리되 코스프레처럼 되면 안 돼. 루피 흉터랑 조로 눈 흉터 같은 건 그대로 가야 하고. 실제 한국 배우 중에서 골라서 누가 무슨 역인지 적어줘. 밀짚모자 일당 여섯 명 캐릭터 컷이랑 극장에 걸 메인 포스터 1장.

Inspect the task →

Website · 2 outputs

롯데 자이언츠 유니폼 새로 디자인해줘. 매년 조금씩만 바뀌어서 지금은 옛날 옷 같아. 요즘 나오는 유니폼처럼 가줘. 세로줄은 빼도 돼. 대신 뺐으면 그 자리를 뭘로 채웠는지 보여줘. 홈, 원정, 서드 3종과 착용 컷 2장. 등번호와 이름 들어갈 자리도 같이.

Inspect the task →

Website · 2 outputs

Build a Y2K-style interactive website where I can create glitter

Inspect the task →

Website · 2 outputs

Create a low-poly 3D island as a .glb model file, plus a set of 5

Inspect the task →