Skip to content

FunAudioLLM

A framework for natural voice interaction with LLMs

Visit official site

What funaudiollm is

The framework is built to make voice interaction between people and large language models more natural.

What you can do with it

  • Recognize speech across many languages
  • Detect emotions and audio events
  • Generate natural multilingual speech

Key features

  • SenseVoice with very low latency
  • CosyVoice zero-shot context generation
  • Cross-lingual voice cloning support

Getting started

  • Visit the GitHub page for model details
  • Pick SenseVoice or CosyVoice and get the code
  • Set up training and inference environments

Pros & cons

✓ What we like

  • Emotion recognition and audio event detection are listed
  • Voice generation supports language, timbre and emotion control

! What to watch out for

  • Users choose between SenseVoice and CosyVoice by need
  • Training and inference environments are set up locally or in the cloud

FAQ

What are the two models?

SenseVoice and CosyVoice are named in the material.

How many languages are supported?

The material states SenseVoice supports over 50 languages.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review