Back: Catalog

VoxCPM

VoxCPM2 is a diffusion-based autoregressive text-to-talk (TTS) synthesis system that does not use discrete markers. The 2 billion parameter model is capable of producing 48kHz studio-quality audio in 30 different languages. The system supports voice design and controlled cloning capabilities over reference audio.

VO

VoxCPM

95

OpenBMB · Voice & Text-to-Speech

STARS40.9K
FORKS4.7K
COMMITS172
CONTRIBUTORS29
VERSION2.0.3
UPDATED23 days ago
PythonTTSSes KlonlamaDifüzyon ModeliApache-2.0

Common commercial products in this category: ElevenLabs, Amazon Polly, Google Cloud Text-to-Speech

What can it do?
  • You can design new sound profiles from scratch using natural language definitions.
  • With short reference audio clips, you can clone sound by controlling emotion and tone.
  • You can convert texts in 30 different languages into direct speech without specifying a language tag.
User reviews
  • • NeutralGitHub Issue

    Voice cutoff at the end of words — Polish language

    “**VoxCPM version:** 2.0.2 --- ### Description When generating speech in **Polish**, the voice occasionally cuts off abruptly at the end of a word. The final syllable or consonant is truncated, making the output sound unnatural. This happens with default generation parameters.…”

Website GitHub

Verified against README · Oct 26, 2026 · source

My note

Alternative pages

ElevenLabs alternativesAmazon Polly alternativesGoogle Cloud Text-to-Speech alternatives

More in: Voice & Text-to-Speech