MegaTTS3
ByteDance's third-gen TTS — 0.45B params, high-quality CN/EN synthesis and voice cloning
ByteDance's third-generation TTS: 0.45B parameters for high-quality Chinese/English synthesis and voice cloning.
Voice AI: lifelike TTS, accurate ASR, and real-time voice agents.
Key features
- Lightweight TTS diffusion transformer with 0.45B parameters
- High-quality voice cloning from a short sample under 24s
- Bilingual Chinese and English with code-switching
- Accent intensity control and fine-grained pronunciation adjustment
Good for
- Researchers cloning voices in Chinese or English
- Apps needing controllable bilingual speech synthesis
Stack & tags
PythonPyTorchGradioDockerttsvoice-cloningbilingualdiffusionlightweightchinese-english
Quick start
conda create -n megatts3-env python=3.10
conda activate megatts3-env
pip install -r requirements.txtFrom the project README on GitHub.