Back: Catalog

CosyVoice

Developed by Alibaba FunAudioLLM, this system is a solution for converting audio from a multilingual text (TTS) based on large language models. The structure, which supports 9 different languages as well as a large number of Chinese dialects, offers the possibility to edit pronunciation over Pinyin and CMU phonemes. Thanks to its bidirectional flow support, it can produce low latency audio up to 150ms level.

CO

CosyVoice

60

FunAudioLLM · Voice & Text-to-Speech

STARS21
FORKS2
COMMITS551
CONTRIBUTORS52
VERSIONv2.0
UPDATED1 yr ago
PythonLLM-BasedMultilingualZero-ShotApache-2.0

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

What can it do?
  • You do zero-shot audio cloning in 9 different languages and 18 more Chinese dialects
  • Using Pinyin or CMU phonemes, you make pronunciation corrections on the text
  • You produce sound by controlling parameters such as emotion, speed, and sound level with instructions
Heads up
  • • Website is unreachable (checked: 2026-10-31)
  • • Last update over 1 year ago
GitHub

My note

Alternative pages

ElevenLabs alternativesAmazon Polly alternativesGoogle Cloud Text-to-Speech alternatives

More in: Voice & Text-to-Speech