Feature Request: Persistent Speaker Profiles (Cross-file Speaker Recognition)

[9stannn]

Hi team, thanks for the amazing work on Buzz!

Currently, the speaker diarization feature works beautifully for isolating voices within a single audio file. However, for users who record multiple files with the same people (e.g., a podcast with recurring hosts, using the exact same microphones and room acoustics), we have to manually rename "Speaker 0", "Speaker 1", etc., for every single new file.

Proposed Feature:
It would be incredibly helpful to have a "Speaker Profile" or "Voice Library" feature.

• The ability to save a voice print/embedding from an identified speaker in one file.
  

• When running speaker identification on a new file, Buzz could compare the detected voices against the saved user profiles.
  

• Automatically assign the correct name if a match is found (perhaps with a configurable confidence threshold).
  

Use Case:
This would save a massive amount of time for podcast editors, researchers, or meeting transcribers who process dozens of episodes with the same recurring speakers. Since the acoustic environment and hardware often remain identical in these scenarios, the voice matching might be highly accurate.

Thanks !

[9stannn]

Just to add another idea to this request: it would be helpful to have a setting to automatically "Identify speakers" either during the initial transcription process or right after it finishes.