Acknowledgements

Built on open work

Family Legacy stands on excellent open-source research. These models help transcribe, search, and organise your recordings — and they all run on our own hardware, so your family's voices and photos never leave to a third party.

• pyannote.audioMIT

  Works out who is speaking in a recording, so a conversation can be attributed to the right person.

• InsightFaceMIT (non-commercial models)

  Recognises the same face across photos, so we can suggest who is in a picture (you always confirm).

• Whisper / faster-whisperMIT

  Transcribes recordings into searchable text, in the language they were spoken.

• nomic-embed-textApache 2.0

  Turns transcripts into searchable meaning, so you can find a story you only half-remember.

With thanks to the researchers and maintainers behind each. See each project for its full licence and citation. How we handle your data.