Skip to content

VideoChat

Real-time voice interaction with a digital human

Visit official site

What VideoChat is

VideoChat is described as a real-time voice interaction digital human project with an end-to-end voice path and a cascaded path, drawing on ASR, LLM, MLLM, TTS and THG.

What you can do with it

  • Customise the digital human figure and voice without training
  • Clone a voice for the digital human
  • Use the end-to-end or cascaded voice path
  • Try the online demo for real-time interaction

Application scenarios

  • Online customer service providing 24-hour consultation
  • A virtual host for news reading and entertainment
  • A virtual teacher assisting in education

Getting started

  • Clone the project code with git clone
  • Configure Ubuntu, Python and CUDA versions
  • Install dependencies from requirements.txt
  • Download the required weight files

Pros & cons

✓ What we like

  • Figure and voice can be customised without training
  • An online demo is listed for real-time trials

! What to watch out for

  • Setup names Ubuntu, Python and CUDA versions
  • Weight files must be downloaded according to the guide

FAQ

Which voice paths are supported?

The material lists an end-to-end path (GLM-4-Voice - THG) and a cascaded path (ASR-LLM-TTS-THG).

Does it need training to change the voice?

The material states the figure and voice can be customised without training, and voice cloning is supported.

Last reviewed: 2026-11-15

More AI digital human tools

View all →

How we review