YE Zhen

icon company Hong Kong University of Science and Technology icon location Hong Kong icon location Speech synthesis, Audio generation, Speech LLM

Results 2 issues of YE Zhen

issue about speaker embedding

Hello, thanks for your great work. https://github.com/svc-develop-team/so-vits-svc/blob/4078cb71a4400ecb158165aa64610269c1a654ec/models.py#L437 I am confused about why not add speaker embedding for z_p prediction. Because the z has used the speaker embedding and z_p should...

not urgent

Code for objective evaluation

I'm confused about the objective assessment such as STOI and PESQ because they need the two input audios to have the same length. However, the predicted audio may have a...