search

pyannote / pyannote-audio

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding
http://pyannote.github.io
MIT License
5.94k stars 755 forks source link

Performing Speaker Verification #651

Closed ajay960singh closed 3 years ago

ajay960singh commented 3 years ago

How can I perform Speaker Verification with Pyannote?

I've an audio segment of the user and I want to compare new instances to this segment to verify the user.

hadware commented 3 years ago

If i'm not mistaken, there isn't any Speaker Verification pipeline in pyannote.

However, it's pretty easy to build one yourself using the speaker embedding pre-trained pipelines. Here's an older issues that might be relevant for your problem: #391

hbredin commented 3 years ago

Also relevant: https://huggingface.co/hbredin/SpeakerEmbedding-XVectorMFCC-VoxCeleb

ajay960singh commented 3 years ago

@hbredin This is giving us reasonable accuracy. Seems like pyannote.audio 2.0 is still in development. When do you expect it to go into production?

hbredin commented 3 years ago

pyannote.audio is a research project, developed by academic researchers like me. Therefore, it will most likely never be production-ready.

Feel free to drop me an email to discuss potential collaborations with your company (e.g. through sponsoring of a software engineer working on making pyannote.audio closer to what you call production).

Closing as the original question has been answered.