PersonaLive - The University of Macau and other open source real-time AI portrait animation generation live framework

What is PersonaLive

PersonaLive is the open source real-time AI face-swapping live streaming framework by the University of Macau,dzine.aiDeveloped in conjunction with the GVC Lab at the University of the Greater Bay Area. It can realize low-latency and high-frame-rate digital human driver on ordinary consumer-grade graphics cards (12GB video memory), and supports real-time control of character expressions and movements in still photos through the camera. The core technology includes hybrid motion control, step less appearance distillation and microblock streaming generation, which significantly improves the inference speed (latency below 100ms) and is suitable for scenarios such as virtual anchor and short video creation.

Features of PersonaLive

• Real-time and Efficient: With just a single 12GB video memory GPU, it is possible to realize real-time, streamable, unlimited-length portrait animations that combine low latency with high quality, with latency as low as 40-50ms per frame.
• low resource consumption: It has low hardware requirements and runs smoothly even on less configured devices, lowering the threshold for use.
• High quality animation generation: The generated animation is of high quality, and the character looks will remain stable without crumbling or going out of shape under long runtime.
• Wide range of application scenariosIt is suitable for live broadcasting, short video production, animation production, virtual anchor, emoticon imitation, role-playing, social entertainment, product demonstration, education and training, advertising and marketing, and many other scenarios.
• Innovative technical architecture: Contains 6 well-designed neural network components with innovative techniques such as mixed motion signals, step less appearance distillation strategy and autoregressive microblock flow generation paradigm to enhance performance and effectiveness.
• usability: Detailed usage and documentation are provided, including steps for environment configuration, weight download, offline and online inference, etc., which makes it easy for users to get started quickly.
• open source: As an open source project, users are free to download, use and modify the code for secondary development and customization.

PersonaLive's Core Benefits

• Real-time and low latency: It can realize real-time unlimited length portrait animation generation with extremely low latency, as low as 40-50ms per frame, which is suitable for scenarios with high real-time requirements, such as live broadcasting and real-time interaction.
• Low hardware requirements: Only a single GPU with 12GB of video memory is required to run it, lowering the hardware threshold and making it available to more users on common devices.
• High quality and stability: The generated animations are of high quality and the characters remain stable over long runs without crumbling or going out of shape.
• Innovative Technology Architecture: Innovative techniques such as mixed motion signals, step less appearance distillation strategy and autoregressive microblock flow generation paradigm are employed to enhance generation efficiency and effectiveness.

What is PersonaLive's official website

• GitHub repository:: https://github.com/GVCLab/PersonaLive
• HuggingFace Model Library:: https://huggingface.co/huaichang/PersonaLive
• arXiv Technical Paper:: https://arxiv.org/pdf/2512.11253

Who PersonaLive is for

• content creator: Suitable for creators who need to quickly generate high-quality animated content, such as short-video producers and live anchors, who can control the virtual image to perform in real time to enhance the fun and interactivity of the content.
• Animation and Film Practitioners: It can be used as an auxiliary tool for animation production, quickly generate animation prototypes or special effects, improve production efficiency, applicable to animation studios, film and television special effects team, and so on.
• Virtual Image Developer: Can quickly generate the dynamic effects of virtual image, suitable for the development of virtual anchor, virtual idol and other virtual image-related projects to meet the individual needs.
• educator: It can be used to produce teaching animation, explain knowledge through virtual image, increase the fun and attraction of teaching, suitable for educational institutions and online education platform.
• game developer: It can be used to generate in-game character animations to enhance the game's visual effects and interactive experience for game development teams.
• Advertising and marketing staff: It can be used to create advertising animation, displaying products or services through virtual images, attracting users' attention and enhancing advertising effects, suitable for advertising agencies and marketing teams.