SRT Whiteboard Animation

srt-whiteboard-animation is an Agent Skill open-sourced by 江哥是老登啊 (GitHub user geeklee) that turns SRT subtitles into a hand-drawn whiteboard video on warm cream paper. It starts by parsing the subtitles and proposing a storyboard and illustration plan at 25 to 35 seconds per scene. Once you approve, your AI tool's own image feature draws minimal line art to a fixed style guide that forbids any text in the picture. It then maps each region of the picture to the moment in the subtitles when it should appear, and opens a local preview board where you can drag regions and adjust their order, timing and linked subtitle line. Finally a Python script renders each scene: a pen tip moves continuously through each region, laying down line art first and colour after, and several scenes can be merged into one MP4. What sets it apart is that the drawing order follows the story in the subtitles, setting first, then characters, then the action and its result, and regions whose turn hasn't come stay completely hidden; SKILL.md also makes the agent stop and wait for your approval after every step, so no render time is wasted before the storyboard is settled.

Who it's for:creators of Chinese-language explainers, story narration, courses or short videos who want a hand-drawn whiteboard look. You'll need to be comfortable with an AI coding tool that runs commands, such as Claude Code or Codex.

Before you start: the README gives no one-line install; you place the whole repo folder in the directory your AI tool reads Skills from, and the repo includes a metadata file for Codex. You need Python. On first run a script creates a separate virtual environment inside the Skill folder and goes online to pip-install opencv-python, numpy, av and Pillow at unpinned versions; MP4 encoding uses PyAV, so no separate FFmpeg install is needed. The line art comes from your AI tool's built-in image generation, at that tool's prices. The preview board opens in a browser; in Chrome or Edge, with your permission, it reads and writes the scene folder you pick, while other browsers need you to download and replace files by hand. SKILL.md requires all explanations, storyboards and interface text to be in Chinese, and neither the README nor SKILL.md mentions adding voice or music to the finished video. MIT licensed.

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