Video Generator

The problem

Producing a polished tutorial video means writing a script, recording narration, adding captions and highlighting clicks by hand.

The solution

ScreenCraft Studio takes an MP4 screen recording (with or without a transcript). A vision model drafts the narration script and finds clicked UI components. You review everything on a synchronized timeline, then export a narrated MP4 with burned-in captions and yellow click highlights.

Highlights

• A vision LLM (Ollama, OpenRouter) drafts the narration script from sampled video frames
• Automatic yellow click highlights with a second component-localization pass and confidence filtering
• Edge TTS narration with voice selection, pace control and word-level subtitle timing
• FFmpeg time-stretching keeps speech in sync without changing pitch
• FastAPI + Redis background worker, encrypted provider keys, and an optional n8n webhook on completion