How to build Tuned-GE2E-SayEN-EarSpeech based on RTVC-7 Voice Cloning Model

How to build Tuned-GE2E-SayEN-EarSpeech based on RTVC-7 Voice Cloning Model? I believe many inexperienced people are at a loss about it. Therefore, this paper summarizes the causes and solutions of the problem. Through this article, I hope you can solve this problem.

0. Description

Almost identical to the version of GE2E in https://blog.csdn.net/u013625492/article/details/114433738

The difference is that the text is in English and trained with English data sets such as VCTK.

1. Git Clone

Https://github.com/ruclion/Tuned-GE2E-SayEN-EarSpeech

two。 Training data 2.1. VCTK

AutoVC has been trained before.

The logic of processing data starts with: / ceph/home/hujk17/Tuned-GE2E-SayEN-EarSpeech/FaPig_extract_GE2E_VCTK_nosli.py

Train, val, unseen. Only use train when training

3. Speaker embedding

At 2.1. Also extracted from the, using GE2E

4. Preprocessing data-> mel

At 2.1. Also extracted from the, using the laboratory mel. ● "∀" ●, it's convenient at last.

5. Change the code

Symbols does not need to move, it retains the English punctuation

The path of train.txt needs to be changed.

Take Kiss as the model name, and the logic starts with Kiss_train.py

Train directly with the original little Batch, batch_size = 12

6. GE2E version of logic

The Embedding of 256is still followed by a FC, which is the same as the Chinese one.

It is called FaPig in Chinese and Kiss in English. The logic starts with Kiss_train.py.

From synthesizer.FaPig_train import train and from synthesizer.Kiss_train import train are actually exactly the same, just to unify the format and make a copy.

After reading the above, have you mastered the method of building Tuned-GE2E-SayEN-EarSpeech based on RTVC-7 Voice Cloning Model? If you want to learn more skills or want to know more about it, you are welcome to follow the industry information channel, thank you for reading!