Script Planner Assistant▸ 
AI Apps 
AI News 10 Excellent Open-Source TTS Voice Models Worth Trying
10 Excellent Open-Source TTS Voice Models Worth Trying
In my usual text-to-speech (TTS) research, there are many excellent text-to-speech (TTS) models worth recommending. Below are 10 outstanding TTS voice models I have compiled that perform remarkably in research and practical applications:
The TTS model used by the AI live-streaming sales large model Streamer-Sales, supporting English, Japanese, and Chinese for zero-shot text-to-speech (TTS).
Integrated toolkit: github: https://github.com/RVC-Boss/GPT-SoVITS
Produces stable and excellent results, supports voice cloning. Trained on 300,000 hours of English, Chinese, and Japanese audio data.
Model: https://huggingface.co/fishaudio/fish-speech-1.2
Online version: https://fish.audio/zh-CN/
3、ByteDance's Seed-TTS (not open-source)
Supports multiple languages including English and Chinese, capable of same-language and cross-language generation, handling various types of text such as narrative text, emotional expression, and descriptive text, and generating corresponding voice based on different emotions and contexts.
Project: https://bytedancespeech.github.io/seedtts_tech_report/
4、ChatTTS
Conversational TTS (multi-speaker capable), supports Chinese and English, with delicate prosody.
github: https://github.com/2noise/ChatTTS
Model: https://huggingface.co/2Noise/ChatTTS/tree/main
5、Parler-TTS
Can control pitch, speed, gender, noise level, emotional characteristics, and more
github: https://github.com/huggingface/parler-tts
Model: https://huggingface.co/parler-tts
6、MetaVoice-1B
Supports multiple languages, achieving emotional speech rhythm in English processing
github: https://github.com/metavoiceio/metavoice-src
7、MARS5-TTS
MARS5 can generate speech for scenarios with complex and diverse prosody such as sports commentary and anime
github: https://github.com/Camb-ai/MARS5-TTS
Natively supports English, Spanish, French, Chinese, Japanese, and Korean, offering flexible voice style control and zero-shot cross-lingual voice cloning capabilities.
github: https://github.com/myshell-ai/OpenVoice
Supports bilingual Chinese and English, with over 2,000 different voice timbres
github: https://github.com/netease-youdao/EmotiVoice/blob/main/README.zh.md
10、StyleTTS2
A text-to-speech library approaching human-level quality, leveraging style diffusion and adversarial training with a large speech language model (SLM) to achieve human-level TTS synthesis.
github: https://github.com/yl4579/StyleTTS2
The above are the 10 outstanding TTS voice models I have compiled. Each of these models has its own strengths and weaknesses across different application scenarios and hardware configurations. Choosing the right model requires evaluation based on the specific application and resources.
