Proven prompts behind viral AI-made videos across 4 models — Claude, Muse AI, ChatGPT, Gemini. Browse by model, category, or tech tag, copy a creator's prompt, adapt it to your own model and renderer. Use when you need a motion-graphic, explainer visual, 3D scene, interactive animation, or AI video-gen prompt that already worked in the wild.
中文简介正在准备中。当前显示原文说明。
安装前可以查看交给 AI 代理的指令内容。
Real prompts behind viral AI-made videos — collected across Claude, Muse AI, ChatGPT, and Gemini. Every entry links the creator's original post so you can see the result and credit the source.
Two genres live here side by side:
| model | entries | mostly |
|---|---|---|
| claude | 475 | code-driven (Opus 5.5) |
| muse-ai | 7 | official video-gen prompts + early creator clips |
| chatgpt | 24 | code-driven + Sora 2 |
| gemini | 39 | Veo 3 video-gen |
Categories (shared across models): explainer (diagram builds, concept visualizations, data-viz), motion (kinetic type, particles, showreel-style motion graphics), 3d (Three.js/WebGL/GLSL scenes), interactive (playable pieces, physics toys).
Top tech tags: canvas, svg, threejs, shader, gsap, css, ai-video, veo, sora.
data/videos.json — machine-readable index (545 entries). Fields: slug, model, model_version, title, author, author_url, post_url, category, tech_tags[], prompt, prompt_partial, added, featured.data/provenance.json — per-entry curation notes (why it's notable, which source table it came from) for the 70 hand-curated entries.prompts/<slug>.md — the prompt text as the creator shared it, one file per entry.data/videos.json by model, category, tech_tags:
python3 -c "
import json
vs = json.load(open('data/videos.json'))
for v in vs:
if v['model']=='gemini' and 'veo' in v['tech_tags']:
print(v['slug'], '-', v['title'])
"
prompts/<slug>.md (or the prompt field). Note prompt_partial: true — the creator only shared part of it. Treat those as inspiration, not recipes.post_url is on every entry.Every entry credits its creator via post_url — the prompts are their words, quoted with credit. "Viral" is survivorship bias: these are the hits. Use them for craft, not as a guarantee.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Master 1Password: vaults, Watchtower, passkeys, SSH agent, CLI, and family/team administration. Use when getting full value from 1Password personally or administering it for others.
中文简介正在准备中。当前显示原文说明。
Create 3D visuals with modeling, texturing, lighting, rendering, and optimization for web and product.
中文简介正在准备中。当前显示原文说明。
Create 3D web experiences: scene setup, models, materials, lighting, animation, scroll-driven scenes, and performance budgets. Use when adding 3D to websites beyond basic demos.
中文简介正在准备中。当前显示原文说明。
Writing abstracts that get papers read — structured content, the 5-sentence core, and journal-specific constraints.
中文简介正在准备中。当前显示原文说明。
Learn effectively from courses and academies: choosing programs, studying actively, and converting courses into skills. Use when investing time/money in structured learning.
中文简介正在准备中。当前显示原文说明。
Audit designs for accessibility with WCAG checklists covering color, type, focus, motion, and content.
中文简介正在准备中。当前显示原文说明。