Kokoro 82M V1.1 Zh

🚀 Kokoro - An Open-Weight Series of TTS Models

Kokoro is an open-weight series of small but powerful TTS models. It aims to provide high - quality text - to - speech capabilities with relatively small model sizes.

🐈 GitHub: https://github.com/hexgrad/kokoro

This model is the result of a short training run that added 100 Chinese speakers from a professional dataset. The Chinese data was freely and permissively granted to us by LongMaoData, a professional dataset company. Thank you for making this model possible.

Separately, some crowdsourced synthetic English data also entered the training mix:

• 1 hour of Maple, an American female.
• 1 hour of Sol, another American female.
• And 1 hour of Vale, an older British female.

This model is not a strict upgrade over its predecessor since it drops many voices, but it is released early to gather feedback on new voices and tokenization. Aside from the Chinese dataset and the 3 hours of English, the rest of the data was left behind for this training run. The goal is to push the model series forward and ultimately restore some of the voices that were left behind.

Current guidance from the U.S. Copyright Office indicates that synthetic data generally does not qualify for copyright protection. Since this synthetic data is crowdsourced, the model trainer is not bound by any Terms of Service. This Apache licensed model also aligns with OpenAI's stated mission of broadly distributing the benefits of AI. If you would like to help further that mission, consider contributing permissive audio data to the cause.

^{[1] LongMaoData had no involvement in the crowdsourced synthetic English data.
[2] The following Chinese text is machine - translated.}

• Releases
• Usage
• Samples ↗️
• Model Facts
• Acknowledgements

✨ Features

• Small but Powerful: Kokoro is a series of small - sized yet high - performance TTS models.
• Multilingual Support: Supports both English and Chinese.
• Diverse Voices: Incorporates voices from different genders and accents.

📦 Installation

You can install the necessary libraries to run the model as follows:

!pip install -q kokoro>=0.8.2 "misaki[zh]>=0.8.2" soundfile
!apt-get -qq -y install espeak-ng > /dev/null 2>&1

💻 Usage Examples

Basic Usage

You can run this cell on Google Colab.

from IPython.display import display, Audio

!wget https://huggingface.co/hexgrad/Kokoro-82M-v1.1-zh/resolve/main/samples/make_en.py
!python make_en.py
display(Audio('HEARME_en.wav', rate=24000, autoplay=True))

!wget https://huggingface.co/hexgrad/Kokoro-82M-v1.1-zh/resolve/main/samples/make_zh.py
!python make_zh.py
display(Audio('HEARME_zf_001.wav', rate=24000, autoplay=False))

Advanced Usage

TODO: Improve usage. Similar to https://hf.co/hexgrad/Kokoro-82M#usage but you should pass repo_id='hexgrad/Kokoro-82M-v1.1-zh' when constructing a KModel or KPipeline. See make_en.py and make_zh.py.

📚 Documentation

Releases

Model Facts

🔧 Technical Details

• Model Architecture: The model is based on StyleTTS 2 and ISTFTNet architectures. StyleTTS 2 provides a framework for generating speech with different styles, and ISTFTNet is used for the inverse short - time Fourier transform in the speech synthesis process.
• Training Data: The training data includes Chinese data from LongMaoData and crowdsourced synthetic English data. The combination of these data sources helps the model learn different language patterns and voice characteristics.

📄 License

This project is licensed under the Apache - 2.0 license.

Acknowledgements

TODO: Write acknowledgements. Similar to https://hf.co/hexgrad/Kokoro-82M#acknowledgements