OpenBMB/VoxCPM
Repository · https://github.com/openbmb/voxcpm
VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning
First seen by Oriel on .
AI summary
VoxCPM2 is a 2B-parameter, tokenizer-free multilingual text-to-speech model that generates 48kHz audio and supports voice design, controllable cloning, and 30 languages.
- Its four-stage pipeline operates in AudioVAE V2 latent space, using a diffusion autoregressive architecture built on a MiniCPM-4 backbone.
- Voice Design creates voices from text descriptions; cloning uses reference audio, with optional transcript-based continuation for closer preservation of vocal detail.
- The project reports real-time factor around 0.30 on an NVIDIA RTX 4090 with standard PyTorch, and around 0.13 using Nano-vLLM.
- Deployment options include Python, vLLM-Omni, Nano-vLLM, and llama.cpp-omni; the latter reports an RTF of about 1.76 on Apple M4 Pro / Metal.
- Weights and code are released under Apache-2.0, and the listed runtime requirements include Python 3.10 to below 3.13, PyTorch 2.5.0 or later, and CUDA 12.0 or later.
#text-to-speech #speech synthesis #voice cloning #diffusion models
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Stars
Total repository stars at each observation.
- GitHub Trending: 40,470, latest observation ; +480 over observed period since
Stars gained (1d)
Daily star gain reported by GitHub Trending. A change compares two daily values; it is not additional stars between observations.
- GitHub Trending: 150, latest observation ; +6 over observed period since
Forks
Total repository forks at each observation.
- GitHub Trending: 4,598, latest observation ; +44 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| GitHub Trending | openbmb/voxcpm | https://github.com/OpenBMB/VoxCPM Read discussion |