Index your video hoard by face, object, and whispered word

What it does Edit Mind watches folders of video files and runs a background AI pipeline to extract transcripts, recognize faces, detect objects and text, and analyze scenes. It stores these multi-modal embeddings in ChromaDB and serves a web interface where you query your library in natural language—“find the shot with the red car and the director speaking”—without touching a cloud API unless you opt into one. The whole stack ships as Docker containers, with a CUDA variant for GPU acceleration.

The interesting bit The project is essentially a polyglot mini-platform: a TypeScript/React frontend and Node.js job queue talk to a Python ML service running PyTorch, Whisper, and friends. The author also sells a desktop app with DaVinci Resolve and Final Cut Pro integration to fund the free self-hosted version—a refreshingly honest way to keep the lights on.

Key highlights

• Runs fully offline via Docker Compose, with local Ollama support for NLP and an optional Google Gemini fallback.
• Indexes by spoken word, face identity, objects, on-screen text, and scene context.
• Vector semantic search through ChromaDB lets you jump to specific moments with plain-English queries.
• Monorepo architecture using pnpm workspaces, React Router v7, Express/BullMQ, and Prisma-backed PostgreSQL.
• A paid desktop edition offers one-click install and direct NLE integration for macOS and Windows.

Caveats

• The README explicitly warns the project is not yet production-ready, with incomplete features and occasional bugs.
• “Fully local” is configurable but not guaranteed: you can route NLP through the Google Gemini API instead of Ollama.
• The self-hosted path still requires Docker comfort; the author built the desktop app partly because users complained about server setup.

Verdict Video editors, archivists, and self-hosting tinkerers with sprawling local libraries should track this closely. If you need a bulletproof, finished asset manager today, wait for v1.0—or preorder the desktop app.