Description
Making knowledge shorts is annoying at both ends: AI video tools charge per run, the picture jitters and one changed sentence regenerates the whole clip, while editing by hand means placing every frame yourself. whiteboard-teacher-video is an Agent Skill that turns a script into a hand-drawn explainer video: shapes are drawn stroke by stroke, text is written letter by letter, and the picture follows whatever the voice is saying.
It works nothing like a video model: every frame is drawn by code, not generated as pixels. That lets the video be cached in segments and rendered in parallel, the same moment renders identically twice, and changing one sentence re-renders only the affected segment in seconds. Rendering happens in your local Chrome with no image-generation API at all; the only possible cost is the voice, and the default edge-tts is free.
Hand-drawn animation: rough.js style lines, five visual styles to choose from.
Voiceover and sound effects: the script becomes narration with sound effects added automatically, in one pass.
Real content in frame: numbers, charts, code, KaTeX formulas, brand logos and real web screenshots can all go into the picture, which is hard for purely generative tools.
Second-level re-render: edits only redraw the affected segments.
Let your agent install it: hand your AI the repo link and it runs the setup, pulling the engine and dependencies. It needs Node 22.19+, Git, a local Chrome or Edge and Python 3.9, about 150MB in total.
Licence note: the rendering engine is AGPL-3.0, called as a separate process, so the videos you make are unaffected, but do not copy its source into your own closed-source project.
It works nothing like a video model: every frame is drawn by code, not generated as pixels. That lets the video be cached in segments and rendered in parallel, the same moment renders identically twice, and changing one sentence re-renders only the affected segment in seconds. Rendering happens in your local Chrome with no image-generation API at all; the only possible cost is the voice, and the default edge-tts is free.
Features
Hand-drawn animation: rough.js style lines, five visual styles to choose from.
Voiceover and sound effects: the script becomes narration with sound effects added automatically, in one pass.
Real content in frame: numbers, charts, code, KaTeX formulas, brand logos and real web screenshots can all go into the picture, which is hard for purely generative tools.
Second-level re-render: edits only redraw the affected segments.
Let your agent install it: hand your AI the repo link and it runs the setup, pulling the engine and dependencies. It needs Node 22.19+, Git, a local Chrome or Edge and Python 3.9, about 150MB in total.
Licence note: the rendering engine is AGPL-3.0, called as a separate process, so the videos you make are unaffected, but do not copy its source into your own closed-source project.
