feat: Add Speaker Consistency Evaluation Feature

[6drf21e]

This PR introduces a new feature for evaluating speaker consistency within the project. It includes scripts for generating test audio and assessing the stability of speakers' voice timbre using the ERes2NetV2 Speaker Recognition Model.

Changes

• Added audio_generator.py: Script to generate test audio files based on text data provided in test_data.yaml.
• Added consistency_evaluator.py: Script to evaluate the stability of speaker voice timbre by calculating the cosine similarity between embedding vectors of audio segments.
• Included test_data.yaml: YAML file containing text data for generating audio.
• Updated README.md to include installation instructions, usage guidelines, and details about command-line arguments for both scripts.
• Added requirements.txt to specify dependencies required for running the scripts.

[6drf21e]

colab test
https://colab.research.google.com/drive/1BzyUvrh1XBtaA6UqPLXiKyhUSitNJsqm

[fumiama]

Thanks for your contribution! This helps us a lot. 🤝
However, this repo aims to provide the ChatTTS package so it is hard to add too many bundles 😂.
If you agree, we want to invite you to contribute to a new repo in our organization, 2noise.
Then, we can link to your awesome work to the README in this main repo.

[6drf21e]

I am happy to accept the invitation to contribute to the new 2noise repository. Thank you for your recognition! @fumiama

[fumiama]

What project name would you like? I will create a repo for you and grant you the full permission.

[6drf21e]

Any name will do, I'm not particular about it. Naming things is always the hardest part for me 😂.

[fumiama]

All right 😂. I will decide by my self.

[fumiama]

And, what license would you like? (I suggest AGPL v3)

[6drf21e]

Sounds good!

[fumiama]

OK.

[fumiama]

I have invited you to https://github.com/2noise/ChatEval
Therefore, I would like to close this PR first.
Feel free to push your codes there! 🎉