Hallo/Hallo3
Developed jointly by Fudan University and Baidu, it features audio-driven functionality along with dynamic backgrounds to achieve effects close to those of film-and-television-grade digital humans.
📋 Info
| GitHub Stars | ⭐ 11.0k Stars |
| License | MIT |
| Language | Python |
| Version | Hallo3 |
| Updated | 2026-05-10 |
📖 Overview
Hallo3 is a high-quality digital human system jointly developed by Fudan University and Baidu. It offers significant improvements over its predecessor in terms of image quality, character movements, and support for dynamic backgrounds—delivering visuals close to film-grade standards. Audio drives facial expressions and head movements, and it allows users to fine-tune the model using their own videos. It is suitable for professional creators who seek high-quality video digital humans.
✨ Features
- Near cinematic image quality
- Support for dynamic backgrounds (non-green screen).
- Audio-driven full-body micro-movements
- Trainable — fine-tune using your own videos
- High-quality content from Fudan University and Baidu Scholar
Advertisement