story-to-handdrawn-video

story-to-handdrawn-video is an Agent Skill open-sourced by GitHub user gnipbao, paired with a Remotion-based renderer, that turns a Chinese story or a set of ordered images into a 3:4 vertical hand-drawn animation. Give it a story and the agent splits it into scenes by sentence, sets up the characters, then calls an image tool to draw a "master page" for each scene, with handwritten Chinese captions and the illustration in the same image; the captions are checked for wrong characters before the pages are imported and rendered. There are two animation modes: the layered mode reveals the text, then a black-and-white sketch, then the color; the page-flip mode keeps each full page intact and curls it from the bottom-right corner to reveal the next. It ships 297 style recipes and 30 theme palettes, with 30 curated styles shown by default and an online style library to browse first. You can also hand it a reference image so the agent can study its lines, colors and materials and draw a new story the same way, and save that style for later in your local project. What sets it apart is that the black-and-white sketch is derived locally from the final color image, so the three layers line up without generating any extra images for the animation. The README says it works with Codex, Claude Code, Kimi Code and other tools that support Skills.

Who it's for:creators making Chinese story videos, picture-book readings and hand-drawn explainer clips. You'll need to set up Node and Python yourself, so readers who never touch a command line will find it hard going.

Before you start: clone the repo, install the renderer with npm ci, then copy the Skill folder from skill-package into your AI tool's skills directory. You need Node.js 20 or later, Python 3.10 or later, FFmpeg (with ffprobe), npm, and Chrome or another browser Remotion supports. New artwork comes from whatever image tool your agent can call; the README notes that "local tool" here means called from the current agent, not an offline model, and image generation counts against that tool's quota. Only if you explicitly choose the API route and set OPENAI_API_KEY do generation requests go straight to OpenAI's API, at OpenAI's prices. The output is a silent H.264 MP4 (720×960 preview, 1080×1440 final, 30 fps), so voiceover and music are added in post. The code is MIT; the style material is adapted from other open-source projects, whose licenses are kept in the repo. The demo's background track, Carefree by Kevin MacLeod, is CC BY 4.0 and needs credit if you reuse it. Using the tool doesn't give you rights to a reference image or someone else's style, so check before using another artist's work as a reference. Remotion, which does the rendering, has its own license terms to review separately.

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