liangdabiao/whiteboard-teacher-video is a Python skill that turns an explainer script into a video where drawings and text appear in sequence. According to the README, it offers five styles and renders in local Chrome or Edge to produce an MP4 video and SRT subtitles. There is no published latest release information.
Instead of generating pixels with an image generation API, this skill builds the screen by drawing shapes and writing text in code. Every moment on screen is defined by a renderAt(t) function, so it can jump to a specific point in time or render in separate segments. It draws vector shapes with rough.js, displays formulas with KaTeX, captures the browser screen with playwright-core, and assembles the video with ffmpeg-static. It is designed so that rendering the same input again yields the same screen, and its distinguishing feature is that editing the script only redraws the affected segments. It is a fitting approach for anyone who wants to produce videos locally without generative image API costs.
That said, the subtitles assume Chinese, and the render engine has to be brought in from the sitor-ai/frametell repository as a sparse checkout. So there is a dependency on that repository, and it is hard to regard as a standalone configuration. The license is listed as NOASSERTION, so the terms of use cannot be confirmed from the information provided. The reproducible rendering approach stands out, but the Chinese subtitle assumption and the external engine dependency narrow the range of applicability.