VoxCPM2
TOKENIZER-FREE TEXT-TO-SPEECH

Speech without a tokenizer in the way.

VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning.

Try the model Read the paper →
"Generated, multilingual, true-to-life."illustrative
0:11 / 0:29
Waveform is illustrative — no audio is bundled in this mock.
MultilingualCreative voice designTrue-to-life cloningTokenizer-free
01

No tokenizer step

Generates speech without an intermediate text tokenizer — fewer moving parts between text and voice.

02

Design a voice

Shape new voices creatively, not just replay a fixed set of presets.

03

True-to-life cloning

Reproduce a target voice with fidelity for multilingual generation.