Skip to main content
Back

Edit Mind Lets You Search Hours of Video Locally Without Cloud

A self-hosted, Docker-based system that indexes home video libraries with Whisper, YOLO, and DeepFace, then lets you search them semantically.

Repo, Full Access
  • Edit Mind is an open-source, local-first video knowledge base with over 1,400 GitHub stars
  • Combines Whisper for transcription, YOLO for objects, and DeepFace for face recognition
  • Semantic search runs on ChromaDB, with PostgreSQL for relational data
  • Choose between local Ollama models or Google Gemini for the NLP step
  • Ships as a Docker Compose stack with an optional CUDA variant for NVIDIA GPUs
  • Paid desktop app available for users who want native Apple Silicon GPU support

Anyone who has ever hunted through hours of footage looking for the shot where a specific person says a specific line knows the pain. Edit Mind is an open-source attempt to fix that problem locally, without shipping your media to a cloud provider. It indexes video libraries with a stack of vision and audio models, then exposes them through a natural language search interface.

The project has crossed 1,400 stars on GitHub and runs entirely in Docker containers. The README is upfront that this is early software in active development, with incomplete features and occasional bugs. The architecture and feature set are still worth a look for anyone building video tooling.

What is actually inside the box

Edit Mind is a pnpm monorepo that stitches together a TypeScript web app, a Node background job runner, and a Python ML service. The pipeline pulls apart each video into multiple signal streams and stores the resulting embeddings in a vector database for retrieval.

  • Search by spoken words, objects, faces, and other extracted signals
  • Runs fully locally on any machine with Docker installed, using AI for metadata extraction and semantic search