OpenBMB Open-Sources VoxCPM, a Next-Gen TTS Engine

Text-to-speech just hit a new level. OpenBMB has open-sourced VoxCPM, a tokenizer-free TTS engine that can invent brand-new voice identities from a text prompt.

It runs on your own hardware, so there are no recurring API fees. That makes it worth a look if you build anything with voice.

What Happened

OpenBMB released VoxCPM as an open-source project. The engine generates 48kHz, broadcast-ready audio with a tokenizer-free architecture, which produces more natural prosody, breathing, and pacing.

VoxCPM also supports prompt-to-voice generation. You can describe a voice in natural language and the model creates a completely new voice that does not already exist. For cloning, it gives you fine-grained control over emotional cadence, tone, and vocal nuance.

The repository is live at https://github.com/OpenBMB/VoxCPM.

Why It Matters

Because VoxCPM is 100% open source, you can host it locally and avoid ongoing API fees. The zero-shot voice design is useful for audiobooks, game characters, virtual assistants, or any project that needs a custom voice without recording one.

If you want to try it, star the repo and test the voice generation on your own machine.