Spaces:
Running
on
Zero
Onnx?
This is a great release! Very impressive and MIT. Looking immediately forward, have you all worked on converting these .pt checkpoints to onnx runtime yet so we can run this anywhere? (non-cuda gpus, cpu-only, no pytorch or big dependencies)
If that's the first time you've heard of onnx - reach out might be able to help you all get better support on all platforms.
its not supporting Arabic ! please we need this to support Arabic language "Saudi Arabia"
I am finding that a local installation will only work with about 10-12 sentences at a time. Any more than that and it produces gibberish. It uses the correct voice, but it sounds like nonsense. When I chunk the text into 8 sentence bundles and then concatenate them into a single wav file, there are random sentences missing. Is there any insight into this? Otherwise fantastic product!