I indexed 669 GB of my GoPro videos using my M1 Max computer and local ML models
Points and comments are a snapshot, not live.
Developer indexed 668 GB of GoPro cycling videos using M1 Max and local ML models for searchable clip extraction.
The author processed 628 GoPro videos (15+ hours of footage) on an M1 Max by extracting frames, classifying them with vision models, transcribing audio, and converting all data to embeddings stored in a vector database. This enabled semantic search across videos, search by screenshot or audio file, and automatic export of matched clips to DaVinci Resolve. The pipeline uses open-source models running entirely locally without cloud uploads.
What commenters are saying
Commenters noted DaVinci Resolve 21 now includes built-in local AI indexing (though without custom face tagging), and highlighted the M1 Max's unified memory architecture (400 GB/s bandwidth) as key to performance. A similar project appeared on HN days earlier. Discussion touched on Docker support for Apple GPU acceleration and the broader appeal of local, privacy-preserving workflows over cloud-dependent tools. Lower-ranked comments diverged into unrelated tangents about adult content detection.