microsoft

/

VibeVoice-1.5B

like 1.22k

Follow

Microsoft 14.9k

I'm working on a conversion. I'll upload them in the near future.

I've been able to convert it, but inference is tricky. I'll keep working on it.

Maybe @Xenova or someone else familiar will create their own solution soon as well.

Thanks @urroxyz for working on the ONNX integration! When you’re done, could you create a PR to the main repo?

Yes, I can do that.

Unfortunately, whenever I try to inference, I receive an unintelligible audio file with no speech sounds. I can't figure out a fix.

If you want, I can upload the conversion files and hope someone else can use them to determine proper inference logic?

You can open a draft PR to share your progress and invite others to collaborate.