OpenBMB/VoxCPM

Project overview

VoxCPM is a tokenizer-free text-to-speech model that outputs 48kHz audio, supports 30 languages, and creates or clones voices from text descriptions or short reference clips. It is suitable for developers and creators who need self-hosted, programmable speech synthesis or voice cloning.

Project type

Model Runtime · Audio & Speech

Use cases

Meeting Notes · Audio & Speech

Deployment

Refer to project documentation

License

Apache-2.0

Best for

• Developers and AI engineers seeking a self-hosted, locally runnable text-to-speech model with configurable inference backends.
• Creators who need to synthesize 48kHz audio or clone voices from short reference clips.

Key capabilities

• Creates a brand-new voice from a natural-language description covering gender, age, tone, emotion, and pace without needing reference audio.
• Clones voice timbre from a short reference clip and allows style guidance to steer emotion, pace, and expression.
• Reproduces vocal nuances by continuing from reference audio and its transcript to preserve timbre, rhythm, emotion, and style.
• Outputs 48kHz studio-quality audio directly via AudioVAE V2 with built-in super-resolution.
• Supports full SFT and LoRA fine-tuning to adapt to specific speakers, languages, or domains.

Limitations and risks

• Outputs from Voice Design and Controllable Voice Cloning can vary between runs, requiring multiple generations to achieve a consistent desired result.
• The model officially supports 30 languages; using it with other languages may require additional testing or fine-tuning.
• The voice cloning capabilities carry a risk of misuse in impersonation, fraud, or disinformation, requiring safeguards and responsible deployment practices.

Getting started

• Install the package using pip, after which pre-trained weights are automatically downloaded. Users can then generate audio via the CLI using commands such as 'voxcpm design' or by integrating the Python API into their scripts.

Evidence and sources

• README: VoxCPM is a **tokenizer-free** Text-to-Speech system that directly generates continuous speech representations via an end-to-end **diffusion autoregressive architecture**
• README: pip install voxcpm
• README: For production multi-tenant deployments, use **[vLLM-Omni](https://github.com/vllm-project/vllm-omni)** — the official vLLM project's omni-modal extension with native **VoxCPM2**…
• README: | **VRAM** | ~8 GB | ~6 GB | ~5 GB |
• README: 🔊 **48kHz High-Quality Audio** — Accepts 16kHz reference audio and directly outputs 48kHz studio-quality audio via AudioVAE V2's asymmetric encode/decode design, with built-in sup…

Reviewed at: 2026-07-17T09:16:56Z