skillcast turns a tutorial screencast into a skill your coding agent loads and executes — with the exact commands that were on screen.
“…and now we'll add the test runner and the testing library.”
npm install -D vitest @testing-library/react
Only one of those runs. Every video-to-text tool reaches for the transcript — the half that loses the executable truth. skillcast reads the screen.
Runs entirely in your browser with tesseract.js. Your video is never uploaded, because there is no server to upload it to.
starting…
Measured against a fixture whose contents are known exactly, degraded on purpose. Recall is exact-match, not similarity.
| 720p terminal, large type | 5/5 |
| Re-encoded at 360p, crf 30 | 5/5 |
| Heavy noise added | 5/5 |
| IDE-style, 22px type amid code | 4/5 |
| IDE-style, 16px type amid code | 4/5 |
| IDE-style, 13px type | 2/5 |
| IDE-style, 11px type | 1/5 |
Compression barely matters. Type size does.
Below roughly 16px the OCR starts dropping commands, and no threshold rescues it — so that limit is published rather than left to be discovered.
If you control the recording: shoot at 1080p and raise the terminal font.
The demo is the same logic, ported. The CLI is faster, handles long recordings, and writes the files where your agent already looks.
pip install git+https://github.com/obeskay/skillcast
brew install ffmpeg tesseract # or: apt install ffmpeg tesseract-ocr
skillcast tutorial.mp4
No API key. No model download. Fully offline.