What GLM-4-Voice is
The model is described as an end-to-end speech system from a Tsinghua University team for real-time Chinese and English voice conversation.
What you can do with it
- Convert continuous speech into discrete tokens
- Turn discrete tokens back into speech output
- Change emotion, tone, speed and dialect by instruction
Getting started
- Clone the repository with git
- Install Python dependencies from the requirements file
- Download the speech models and tokenizer
Pros & cons
✓ What we like
- The material lists emotion, intonation and dialect control
- Pretraining used millions of hours of audio
! What to watch out for
- Running it starts from cloning the repository
- A model server script is part of the documented setup
FAQ
Which languages are supported?
Chinese and English speech are named.
How is it started?
Clone the repository, install dependencies, download models and run the model server script.
Last reviewed: 2026-11-15
More AI audio tools tools
View all →-
Accent Guesser Accent recognition from a recording Free tools Creator tools Audio Speech to text Education tools -
AccurateScribe.ai Transcription for audio and video Free tools Creator tools Video to text Audio Speech to text -
Adobe Podcast Browser podcast recording and cleanup Free tools Creator tools Audio Speech to text Audio editing -
Agilotext French-first transcription and summaries Audio Speech to text Podcast Meeting transcription Meeting notes -
AI Dubbing Short video dubbing without signup Free tools Video translation Audio Voiceover Content localization -
AI Mastering Automatic online mastering Free tools Freemium tools Audio Podcast