Skip to content

LiteAvatar

Audio-driven real-time 2D chat avatars

Visit official site
Category Chatbots

What LiteAvatar is

The model is described as an audio-driven real-time 2D avatar generator built for live chat scenes. It pairs efficient speech recognition and mouth shape prediction with a lightweight 2D face model, and it is open source and free.

What you can do with it

  • Extract audio features with an efficient ASR model
  • Predict mouth shape parameters synced with speech
  • Render 2D avatar mouth motion in real time
  • Run inference at 30fps on CPU-only devices

Application scenarios

  • An online education platform offers a real-time virtual teacher avatar
  • A virtual live platform generates a real-time avatar for a streamer
  • An internal video meeting system uses avatars for participation

Getting started

  • Prepare sample data and unpack it to the given path
  • Set up a Python environment with version 3.10 recommended
  • Run the lite_avatar.py script with a data directory
  • Save the inference result as an MP4 video file

Pros & cons

✓ What we like

  • A lightweight 2D face model is friendly to mobile devices
  • Complete code and documentation support integration

! What to watch out for

  • The named output is audio-driven real-time 2D avatars
  • The 30fps figure is stated for CPU-only devices

FAQ

What drives the avatar?

Audio drives it through feature extraction with an ASR model and mouth shape prediction.

What hardware is described?

The material states 30fps real-time inference on devices that use only a CPU.

Last reviewed: 2026-11-15

More AI chatbot tools

View all →

How we review