StyleTTS2 Add Replicate demo

Hi @yl4579 , I accidentally discard the changes and closed the previous pull request. The issue is now fixed and I updated the Replicate demo as well!

Nov 27 '23 10:11 alaradirik

@alaradirik awesome work with the Replicate demo. Any idea as to how to add training to the Replicate implementation?

Dec 06 '23 08:12 platform-kit

@platform-kit sorry for the late reply! I'd be happy to add a train endpoint to the Replicate demo

Dec 21 '23 08:12 alaradirik

@alaradirik Amazing, thank you so much, and happy new year!

Jan 04 '24 08:01 platform-kit

HI @alaradirik. Thank you for your demo. I have a discussion/question that I was hoping I could get your help with #183.

I've been using replicate to train my model on audio with 50 diffusion steps and it still sounds like the robotic default voice. Is my reference_file being recognized? Any help would be appreciated.

output = replicate.run( model_id, input={ "beta": 0.3, "reference_file": "training_audio.WAV", # also tried: "training_audio.mp4" "training_audio.mp3"

"seed": 2, "text": text_to_synthesize, "alpha": 0.3, "diffusion_steps": 50, "embedding_scale": 5, } ) print(output)

Jan 08 '24 05:01 scoroman

Hi @alaradirik ! I know you're busy with Replicate stuff -- have you had any time to dig into adding the training script?

I'm super excited to work with this lib on Replicate. If a donation would help you find some time for this, let me know how much you need. Thanks!

Jan 26 '24 19:01 platform-kit