GitLab Transcend - Intelligent orchestration
Mini toolsPersonaLive
By GVC Lab
PersonaLive is a portrait animation framework focused on live, continuous generation rather than offline short-video synthesis. The repository describes it as a real-time and streamable diffusion framework capable of infinite-length portrait animation, with released inference code, configs, pretrained weights, WebUI support, and an offline streaming strategy that can generate long videos on 12 GB VRAM. It is positioned for expressive live-streaming avatars, with separate offline and online inference pipelines and optional TensorRT acceleration.
Multimodal
Gen 3
Overview
PersonaLive is a real-time diffusion model for expressive portrait image animation in live streaming. It is built to turn a single portrait into continuous, streamable talking-head motion driven by input video, and is specifically designed for infinite-length generation rather than short fixed clips.
About GVC Lab
GVC Lab (Generative & intelligent Visual Computing Lab) is an AI research laboratory at Great Bay University in Dongguan, China. Led by Prof. Xiaodong Cun, the lab focuses on animation, interactivity, intelligence, and visual representation — using generative AI to advance the movie and game industries. Research spans video generation, 3D Gaussian splatting, portrait animation, and video anomaly understanding.
View Company ProfileTools using PersonaLive
No tools found for this model yet.
Last updated: May 5, 2026
