Skip to main content
All projects

Tutorial video generator · Media

Walkie

Recording a product tutorial by hand means retakes, a video editor, and doing it all again for every language and every UI change. Walkie turns one rough walkthrough — “umm”s and mistakes included — into a tight spoken script, a natural voiceover (a stock voice or your own), a video cut to that voice with a zoom and outline on every click, and a matching help article with screenshots. Instead of laying a voice over the recording, it fits each scene's footage to its sentence: a click lands just after its sentence starts, slow stretches are sped up or cut, and the screen holds still when the voice needs longer — so translations stay in sync too. Open source and self-hosted; recordings stay on your machine and only the Gemini calls leave it.

1 / 2

Highlights

  • Chrome extension records clicks and optional narration — or upload an existing screen recording and Gemini finds the clicks
  • Video is fitted to the voice scene by scene: zoom and outline on every click, never sped past 3×
  • Blurs passwords, emails and other personal data for the whole video
  • Add a language and it translates the article and script, re-records the voice, re-times the video and adds captions
  • Edit the video as text — change a sentence and only that scene is re-recorded and re-rendered
  • Export for docs: one zip of Markdown, images, MP4 and captions for every language

Tech stack

  • Python
  • FastAPI
  • Gemini
  • ffmpeg
  • Chrome Extension
  • Docker