What it does Edit Mind is a Dockerized stack that watches a folder of videos, runs them through Whisper (transcription), YOLO/DeepFace (object and face detection), and scene analysis, then dumps the embeddings into ChromaDB. You search the results with natural language through a web UI. Think of it as a private Google Video AI that runs on your NAS.

The interesting bit The project straddles two worlds: a free, Docker-only self-hosted version and a paid desktop app with native Apple Silicon GPU support and direct integration with DaVinci Resolve and Final Cut Pro. The author is blunt about the split—Docker can’t touch Apple GPUs, so the commercial build exists for editors who want speed without terminal gymnastics.

Key highlights

• Multi-modal indexing: transcription, face recognition, object detection, text-in-frame, and scene analysis
• Choice of LLM backend: local Ollama (Qwen 2.5 by default) or Google Gemini API
• Background job queue via BullMQ/Redis; web layer is React Router v7 + Vite
• One-command Docker Compose setup with optional CUDA compose file for NVIDIA GPUs
• Desktop app preorder available for macOS/Windows with lifetime license

Caveats

• Explicitly not production-ready; README warns of incomplete features and bugs
• Requires manual Docker file sharing configuration and two separate env files
• Default admin credentials (admin@example.com / admin) are, well, default admin credentials
• License is a custom “Edit Mind License,” not a standard OSI-approved license

Verdict Worth a spin if you have a sprawling video archive and trust issues with cloud AI. Video editors eyeing the Resolve/FCP integration should wait for the desktop app to ship; everyone else should expect some assembly required.