search

yl4579 / StyleTTS2

StyleTTS 2: Towards Human-Level Text-to-Speech through Style Diffusion and Adversarial Training with Large Speech Language Models
MIT License
4.48k stars 355 forks source link

Is it possible to make onnx model support? #117

Open orkars opened 7 months ago

orkars commented 7 months ago

This is coolest model that I saw. Thank you for this prefect work. For using this to different platforms, is there any supports or information to make inference?

81549361 commented 7 months ago

I also want to know if it can support onnx model

yl4579 commented 7 months ago

You may refer to #39 if you just want portability. I'm not familiar with Onnx so it probably needs to be be done by someone more familiar with this.

adrianlyjak commented 1 month ago

Someone seems to have done this. I'm not sure what hexgrad is, whether its a researcher/hobbyist or business.

https://huggingface.co/hexgrad/styletts2

https://hexgrad.com/

zolero commented 1 month ago

I'm working on this as well, I'll keep you guys posted. You could increate interference / generate speed by quantifying the models and loading in another backend architecture like tensorrt

logikstate commented 2 weeks ago

So who is the mysterious hexgrad? and do they plan to share their most excellent onnx implementation with the rest of us mere mortals? :)