Songting

1. 多音字转音素目前由`jieba`和`cn2an`这两个库进行，但是暂无对此进行改进的打算。如果您希望，可以：1) 输入同音字 2) 在您的应用中修改源代码并自定义g2p 2. 受限于模型架构，不能。但是有一个取巧的方法是将音频prompt慢放之后再输入模型，模型会尝试使用相同的语速（暂未测试该方案）

It must be in Python 3.10

I apologize that I may not have enough availability to implement and test fine-tuning feature recently, but I'll try to find some time to work on it if possible. Whether...

I apologize that I'm not familiar with the usage of Gradio Client API. Please figure it our yourself.

1. 删 2. 我要是知道就好了

Hi there, Of course, you don't need to train Encodec during finetuning. Please simply follow lifeiteng's repo, whose link has been given in read me. Please prepare enough data if...

Thanks for your interest and kind words about this repo. I apologize that I am not an expert in ML programming on AMD GPUs (because I don't have one😥), so...

`ComplexFloat` seems to be a common problem reported by several Mac users, but I personally don't have a MacBook to do debugging. So, I apologize that currently I'm unable to...

Adding language embedding to acoustic tokens doesn't make sense at all. I tend to believe this is a typo error > Why are language embeddings being added to the phoneme...

I'll suggest you to take a look at MetaAI's SeamlessM4T. Although looks promising, direct dubbing is no better than ASR + NMT + TTS.