About GitSkit.
Cut. Clip. Command. A local-first video transcript studio for creators who move fast.
What is GitSkit?
GitSkit is a self-hosted studio for building video supercuts from word-level transcripts. You point it at a folder of videos, it transcribes them using xAI's speech-to-text API, and then you can search, remix, and render clips with surgical precision.
Whether you want to find every time a speaker said "actually," build a sentence they never said, or turn their voice into a melody, GitSkit handles it fast, locally, and without sending your videos anywhere.
Core features
- Supercut Clip every match of a phrase, regex, or AI-generated pattern
- Sentence builder Pick one clip per word to build anything from scratch
- Songify Detect pitch and arrange clips into a melody
- Transcription Word-level transcription via xAI speech-to-text
- Stats Top words, bigrams, and trigrams across all buckets
- Clip library Browse, preview, and download every rendered output
Built for speed
GitSkit was built on top of Wes Bos's Syntax Supercut Studio, reimagined with a faster interface, multi-chain Solana wallet authentication, and a design system built for extended creative sessions.
Everything runs locally. Your videos, your transcripts, your renders. The only external service is xAI for speech-to-text, and that's optional if you bring your own transcripts.
Technology
- SvelteKit for the full-stack framework
- FFmpeg for all video operations
- xAI speech-to-text for word-level transcription
- aubiopitch for melody pitch detection
- Solana Wallet Standard for multichain wallet authentication
- Vercel for cloud deployment