SCD pipeline for speaker change reporting in real time

[Yagna24]

Hi, I was wondering if there's a possibility to output the speaker change time stamps while processing the audio.

For example,
If there's a 4 minute audio and there's a speaker change at Minute 1 i.e. Interval[0,1], is it possible to fetch output or print the start and end timestamps while it goes to the next speaker segment i.e. from Minute 1 to Minute 2 per say (1,2].

I would extremely appreciate your help.

[hbredin]

As of today (Feb. 2022), real-time processing is not supported in pyannote.
I recommend you use @juanmc2005's diart toolkit for this purpose.

[Yagna24]

Thank you for the response.

[Anticope12]

Hi @hbredin , Just reopening this question If there is a plan to process real-time processing in pyannote?

[hbredin]

• Short term: nope, as it would most likely just be a re-implementation of diart (that internally relies on pyannote.audio models anyway).
• Long term: maybe.