Real-time
#3
by
anticope
- opened
Hello team,
Thank you for this project. just wondering how it can process Diarization and transcription real time ? or a delayed of a few sec?
Hi,
Thank you for your interest,
I think we can upgrade this project to another level.
One of my ideas is that instead of collecting all embedding of each segment speaker and clustering offline, we can compare embedding speaker segments in real-time with the previous one to assign speaker labels.
Of course, you need to do some tricks to improve the computing cost.
I hope it could help you and look forward to your application.
Best regards,
Chien
vumichien
changed discussion status to
closed