Awesome-ChatTTS is the officially recommended collection of ChatTTS resources. Feel free to recommend or self-recommend in the issues.

If you find this project helpful in using ChatTTS, please give it a ⭐️ to show your support.

• Official Introduction
• Quick Experience
• Popular Branches
• Interface Description
• Tone Control
• Beginner Tutorial
• FAQs

Project	Star	Highlights
jianchang512/ChatTTS-ui		Provides API interface for third-party applications
6drf21e/ChatTTS_colab		Supports long audio generation and role reading
lenML/ChatTTS-Forge		Voice enhancement and background noise reduction, allows additional prompt words
CCmahua/ChatTTS-Enhanced		Support for Batch File Processing and Exporting SRT Files
HKoon/ChatTTS-OpenVoice		Integration with OpenVoice for Voice Cloning

Project	Star	Highlights
6drf21e/ChatTTS_Speaker		Tone Role Tagging and Stability Evaluation
AIFSH/ComfyUI-ChatTTS		ComfyUi workflow node
MaterialShadow/ChatTTS-manager		Tone Management System

• 1. Input Text: Text to be converted, supports mixed Chinese and English
• 2. Refine text: Whether to process the text into colloquial form
• 3. Text Seed: Configure text seed value, different seeds correspond to different colloquial styles
• 4. 🎲: Generate random text seed value
• 5. Output Text: Text generated after colloquial processing

• 6. Timbre: Preset timbre seed value
• 7. Audio Seed: Configure audio seed value, different seeds correspond to different tones
• 8. 🎲: Generate random audio seed value
• 9. Speaker Embedding: Timbre code, see Tone Control

• 10. Temperature: Control audio emotional fluctuation, range is 0-1, the larger the number, the greater the fluctuation
• 11. top_P: Control audio emotional relevance, range is 0.1-0.9, the larger the number, the higher the relevance
• 12. top_K: Control audio emotional similarity, range is 1-20, the smaller the number, the higher the similarity

• 13. DVAE Coefficient: Model coefficient code
• 14. Reload: Reload model coefficient

• 15. Auto Play: Whether to automatically play after generating audio
• 16. Stream Mode: Whether to enable streaming output
• 17. Generate: Click to generate audio file
• 18. Output Audio: Audio generation result
• 19. ↓: Click to download audio file
• 20. ▶️: Click to play audio file

• 21. Example: Click to switch example configurations

After testing, specifying a timbre seed value each time generates spk_emb and reusing pre-generated spk_emb results in significant differences. It is recommended to prioritize using .pt timbre files or timbre codes (in string form).

In the ChatTTS_Speaker project, preliminary tagging and stability evaluation of timbre seeds have been conducted. Examples can be used to quickly select suitable timbres.

When using the official WebUI, you can directly copy the timbre code and replace the value in 9. Speaker Embedding to achieve timbre control.

When using in a Python script, refer to the compression scheme in issue#07 to implement timbre control.

spk = torch.load("asset/seed_1332_restored_emb.pt", map_location=torch.device('cpu')).detach()
spk_emb_str = compress_and_encode(spk)

params_infer_code = ChatTTS.Chat.InferCodeParams(
    spk_emb= spk_emb_str,  # add sampled speaker
    temperature=.0003,  # using custom temperature
    top_P=0.7,  # top P decode
    top_K=20,  # top K decode
)

With recent iterations, the issues in the source repository code have been mostly resolved. If you encounter problems, it is recommended to first review the Official Documentation in Chinese.