Prompt Motion Lab

Prompt Motion Lab is a bilingual "code video lab" maintained by nonggde. The README says it keeps brief understanding, visual direction, original prompts, timed storyboards, scene code, sound sync and acceptance records on one reproducible line of work, requires no particular model, and does not pass off web templates as a cloud video generator. The repo holds 96 original bilingual recipes in four directions, product flows, data and algorithms, 3D structure and playable demos; a few runnable example scenes driven by time, where one piece of code previews, extracts frames and exports; and an installable skill called prompt-to-video. With it installed, you tell the agent your goal, audience, assets and length; it proposes a visual direction, writes bilingual prompts and a storyboard, builds a code sample, inspects the actual frames and video metadata, and delivers source, MP4, preview and an experiment record. What sets it apart is that only work that has actually been rendered and checked is marked validated; unverified ideas never get labelled as finished.

Who it's for:People who want explainer or product-demo shorts rendered from code; some Node.js familiarity helps. If you don't code, the online lab is a good place to browse the recipes.

Before you start: The online lab runs in the browser with nothing to install; local runs and rendering need Node.js 20 or newer and npm ci. The skill installs into skill-aware AI tools with npx skills add. The interface, recipes and skill are in Chinese and English. The skill states that it will not publish, contact anyone, buy services or claim a model has validated the result. Code, original recipes and the project's own media are free under the MIT license; fonts carry their own licenses.

Based on the project's public documentation. Not tested by AI Inspiration World.