Vinci Clips

Turn a long video into short vertical clips with captions. Self-hosted, open source (AGPL-3.0), runs on your machine with one Gemini API key.

If this saves you an Opus Clip subscription, please star the repo. It is the main way other people find it.

Quick start

git clone https://github.com/tryvinci/vinci-clips && cd vinci-clips && npm run install:all && cp backend/.env.example backend/.env && cp frontend/env.example frontend/.env.local && npm start

Then put your key in backend/.env (GEMINI_API_KEY=...) and open http://localhost:3000.

You need Node.js 18+ and FFmpeg on your PATH. A free Gemini API key works: https://aistudio.google.com/apikey

What it does

• Upload or import. Drag and drop a video up to 2 GB, or paste a YouTube URL.
• Transcribe. FFmpeg pulls the audio, Gemini returns word-level timestamps with speaker labels. Model is set by LLM_MODEL (default gemini-2.5-flash in .env.example).
• Find the clips. Gemini reads the transcript and suggests segments worth posting.
• Cut them. FFmpeg renders the suggested clips to backend/uploads/clips.
• Reframe to 9:16. Face detection (face-api.js) picks a vertical crop window around the speaker.
• Burn captions. Five caption styles: Bold Center, Neon Pop, Typewriter, Bubble Style, Minimal Clean. Rendered through SRT and FFmpeg.
• Streamer layout. Split a gameplay recording with a webcam into a stacked vertical reel.
• Local storage. Transcripts live in backend/storage/db.json, videos in backend/uploads. No database to run.

How it works

Next.js frontend (3000)  ->  Express API (8080)  ->  FFmpeg + Gemini API
                                   |
                     backend/storage/db.json, backend/uploads/

Path	What is there
backend/src/routes/	upload, import, transcripts, analyze, clips, reframe, captions, streamer
backend/src/localdb.js	JSON file store used by the Transcript model
frontend/src/app/	upload page, transcript list, transcript detail
frontend/src/components/	ReframeModal, CaptionGenerator, StreamerGameplayCrop, SubjectDetection

Configuration

Variable	File	Default	Notes
GEMINI_API_KEY	backend/.env	none	Required
LLM_MODEL	backend/.env	gemini-2.5-flash	Any Gemini model with audio input
PORT	backend/.env	8080	Backend port
NEXT_PUBLIC_API_URL	frontend/.env.local	http://localhost:8080	Where the frontend finds the API

Docker: docker-compose up --build. See docker-setup.md.

Privacy and cost

Your video files stay on your disk. The extracted audio and transcript text are sent to the Gemini API for transcription and clip selection, so Google's API terms apply to that data. Cost is whatever Gemini charges for those calls. There is no Vinci account.

Related: Vinci

Want to generate video, not just cut it? Vinci is our MIT-licensed studio that runs image, video, voice, and lipsync models from many providers in one local app, with a price shown before each job. It also has a podcast-to-shorts workflow. If that sounds useful, give it a star too.

Contributing

Issues labeled good first issue are a good place to start. See CONTRIBUTING.md.

License

AGPL-3.0. If you run a modified version as a service, you must publish your changes. For other licensing, email [email protected].

---

If Vinci Clips is useful to you, a star helps a lot.