19beta transcription issue with speaker ID

Using 19 beta I can get the tool to work and create the text, which I can then use to manipulate etc (just as I could in 18.6).

However, when I turn on the new speaker identification prior to running the analysis it fails every time, hanging up and usually crashing resolve.

Is this a problem others are having and is there any suggested remedy?

My hardware should be more than adequate to cope- Ryzen 5800x3d and nvidia 3090 GPU, 64 RAM etc..

Not happening here, I think you should publish a log.

I ran a brief test with four people speaking. No crashes, but all four people were identified as Speaker 6.

Not sure what happened to 1 through 5. Or how it conflated one male voice with three female voices.

Studio 19b1 on Windows 11.

[Shrinivas Ramani]

Hi all

Would you be able to provide some links to sample media for us to look at? Thanks for listing system details in your signatures. It's useful.

Thank you
Shrinivas

I ran a test with three female voices, but a dramatic scene with music and some screaming.
They were identified as separate speakers, but not always the right one. In a quiet scene, identification was near perfect. I'd say, that's where we stand with AI, and it's already much better than 18.

I actually have the game drivers (not studio drivers) installed at the moment. Would that make ANY difference for this particular issue?

Would that make ANY difference for this particular issue?

That would surprise me.

SOLVED - So I updated to the Studio Drivers and it now works fine.

Makes no real sense to me but I'm not complaining.

Since the neural functions run on the GPU, that is absolutely possible.