GitHub - OpenBMB/VoxCPM: VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning

VoxCPM2 is a 2B-parameter, tokenizer-free Text-to-Speech system using a diffusion autoregressive architecture. It supports 30 languages, voice design from natural-language descriptions, controllable voice cloning, and outputs 48kHz studio-quality audio. Trained on over 2 million hours of multilingual speech data, it achieves state-of-the-art or competitive results on multiple TTS benchmarks. The model is fully open-source under Apache-2.0, supports SFT and LoRA fine-tuning, and can run in real-time with RTF ~0.13 on an RTX 4090 via Nano-vLLM. A Python API, CLI, and web demo are provided for quick integration.