Home › Audio Models › CosyVoice

CosyVoice

A multi-lingual large voice generation model which provides full-stack capabilities for inference, training, and deployment of high-fidelity voice synthesis.

Key Information

  • Category: Audio Models
  • Source: Github
  • Last updated: January 09, 2026

Structured Metrics

No structured metrics captured yet.

Links

Visit official website

Canonical source: https://github.com/FunAudioLLM/CosyVoice

Related Tools

Bark

Bark is a transformer-based text-to-audio model by Suno that generates highly realistic, multilingual speech as well as music, background noise, and simple sound effects. It also produces nonverbal cues like laughing or sighing. The model is provided for research purposes with pretrained checkpoints available for inference.
Bark

Bark is an open-source, transformer-based generative audio model by Suno that converts text prompts into realistic, multilingual speech as well as other audio outputs (e.g., music, background noise, and nonverbal cues). It is designed for research and commercial use, offering fast inference on both GPU and CPU.
Chatterbox

A state-of-the-art open source text-to-speech tool featuring imperceptible neural watermarks for secure audio generation.
Chatterbox TTS

Chatterbox TTS is Resemble AI's first production-grade open source text-to-speech model. It offers speech generation with voice cloning and unique features such as emotion exaggeration control, alignment-informed inference, and built-in imperceptible watermarks. It is built on a 0.5B Llama backbone and benchmarked against leading closed-source systems.
coqui/XTTS-v2

A text-to-speech (TTS) voice generation model that enables high-quality voice cloning and cross-language speech synthesis using just a 6-second audio clip. It supports 17 languages, offers emotion and style transfer, improved speaker conditioning, and overall stability improvements over its previous version.
CSM (Conversational Speech Model)

CSM is a conversational speech generation model by SesameAILabs. It generates RVQ audio codes from text and audio inputs using a Llama backbone for language processing and a specialized audio decoder to produce Mimi audio codes, enabling interactive conversational speech synthesis.