People Recognition and Transcripts

Hello all. I started today as an editor on a documentary. Literally day one of the editing process. Lots of reading on the subject matter and starting to organize the media. I am remote editing from proxies. Tons on interviews to go through. Using Resolve 20.
I want to start by recognizing people and transcribing all interviews.
I am finding out that Resolve fails to recognize people. It finds a few and gets no results whatsoever with many. All interviews are mostly closeups with faces filling a significant part of the screen. Any idea if this is happening because I am editing from proxies?. Maybe a bug with v20?

Transcriptions are working great except I am finding that the AI engine is sometimes recognizing multiple people in the interview and tagging them as numbered "Speakers". Many times it labels the same person as multiple speakers. I then assign the correct speaker to all the transcribed dialogue but I end up with lots of empty speaker names that have no dialogue. I have a lot of interviews to go through and the number of empty speakers is growing and I can't find a way to delete the empty speakers from the AI database. Any suggestions?. Might this also be a bug in v20?.

Any help would be greatly appreciated.

Honestly, I would go back to the latest version of 19 for any real work. I find resolve usually needs a few updates on new releases before everything is working properly.

Honestly, I would go back to the latest version of 19 for any real work. I find resolve usually needs a few updates on new releases before everything is working properly.

Thanks for your answer but I am hoping we can make v20 work because a big part of our workflow will depend on transcripts and their exports containing timecode and speaker data.