Documentation gets old fast.
Product videos get old even faster.
I wanted to explore a workflow where an agent can inspect an app, record browser flows, generate narration and assemble videos automatically.
That became Video Docs Builder.
Repo:
https://github.com/tecnomanu/video-docs-builder
The idea is simple:
Playwright records the browser.
TTS generates narration.
FFmpeg assembles the final video.
The output is an mp4 with narration synced to browser interactions.
The project can also generate a small React docs site with the videos embedded.
Why this matters:
- onboarding flows change often
- QA teams need repeatable walkthroughs
- product teams need quick demos
- internal tools need documentation too
- agents can help keep docs closer to the real app
This is not meant to replace human explanation.
It is meant to remove the boring first pass: record the flow, describe the steps, build the video, repeat when the UI changes.
The stack:
- Playwright
- Piper, ElevenLabs or OpenAI for TTS
- FFmpeg
- Vite and React for optional docs site
I am interested in feedback from people maintaining product docs.
Would you trust generated video docs if the source flows were committed and repeatable?
Top comments (0)