CosyVoice

Free

Multilingual zero-shot voice generation from Alibaba.

Open-SourceMultilingualVoice Cloning

Overview

CosyVoice is Alibaba’s open multilingual TTS family offering zero-shot cloning, cross-lingual synthesis and fine-grained emotion/style control. CosyVoice 2 adds low-latency streaming for real-time use. It supports many languages and is well suited to production pipelines.

What makes CosyVoice special?

Focus: Streaming multilingual synthesis.
Availability: Free & Open-Source.
Use cases: Cross-lingual cloning and streaming applications.

Visit official site

Related free models

Chatterbox

A lightweight, fast TTS model built on LLaMA.

Dia

A 1.6B parameter TTS model from Nari Labs.

Kokoro

An 82M parameter TTS model by Hexgrad.

Back to directory