Skip to content

Features

A complete, private library — built from the pixels up.

Six pillars that turn a folder of files into something you can actually search, browse, and rediscover. All of it local.

Semantic search

Search your library the way you remember it — in words, not filenames. Type "golden hour on the beach" or "person in a red coat" and the local vision model finds the frames that match the pixels, ranked by relevance. No manual tagging required to get started.

  • Natural-language queries
  • Visual similarity ("more like this")
  • Combine with filters: person, place, rating
Browse · semantic search Interface preview

Faces & people

Recognize and group every appearance of a person across your whole library. Face embeddings are computed and stored locally — there is no cloud face database, and nothing about who is in your photos ever leaves your machine.

  • Local face clustering
  • Name people once, find them everywhere
  • Private by design — vectors stay on disk
People · face clusters Interface preview

Places & events

Turn the location and time data you already have into trips, events, and places. Browse by where and when without any map provider tracking your queries.

  • Cluster by location and time
  • Trips and moments, automatically
  • No third-party map calls
Places · located library Interface preview

Video scenes & transcripts

Video is a first-class citizen. Clips are split into scenes, keyframes are extracted, and speech is transcribed locally — so you can search inside your footage as easily as your stills.

  • Automatic scene detection
  • Keyframe posters per scene
  • Local speech-to-text transcripts
Videos · scenes & posters Interface preview

Auto-tagging & captions

Every asset is described for you. Local models assign structured tags and write a natural-language caption, giving you a browsable, filterable index of your entire library on day one.

  • Structured, categorized tags
  • Written captions per asset
  • Fully offline vision models
Inspector · tags & captions Interface preview

Personalization

The library learns your taste. As you keep and reject frames, it adapts its ranking to surface the shots you'd actually pick — a model that works for you and stays on your machine.

  • Learns your keep / reject signal
  • Better picks over time
  • Your preferences never sync to a server
Training · keep / reject Interface preview

See it on your own library.

Download