Speaker-Diarization using Stumpy

[OriginalGoku]

Hello everyone,

I have been using the pyannote/speaker-diarization package and have encountered some challenges when attempting to create a production-grade solution with this package.

The primary issue is the inability to "store" a speaker's voice signature, which would enable the system to recognize the same speaker in future extractions.

Another significant concern is scalability. When processing several terabytes of audio files, the computational costs become substantial.

I reached out to @seanlaw to inquire if anyone has attempted to implement speaker diarization using STUMPY. While he couldn't point to a specific project, he reminded me of the Finding Conserved Patterns Across Two Time Series Toturial
It appears that STUMPY could potentially be an effective tool for identifying speakers or capturing unique voice signatures.

I am interested in hearing from anyone who has conducted research in this area.

Thank you.