fudan-generative-vision/hallo2

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

3734 stars · 542 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Hallo2 is a Python research library from Fudan University that animates a single portrait image using audio input, producing long-duration (up to an hour) and high-resolution (up to 4K) talking-head videos. It was published at ICLR 2025 and ships with pretrained models on Hugging Face.

Use cases

• generate a talking-head video from a photo and an audio clip
• animate a portrait image to match speech audio
• create lip-synced avatar videos in 4K
• make long-duration digital human presentations from one image
• produce talking avatar videos for virtual presenters
• research audio-driven portrait animation

When to choose

• you need long (minutes to an hour) high-resolution talking-head videos from a single portrait
• you want a state-of-the-art open-source audio-driven animation model with pretrained weights
• you have GPU resources and are comfortable running Python research code

When to avoid

• you need real-time animation or live streaming avatars
• you lack a capable GPU or want a lightweight consumer app
• you need simple lip-sync on existing video rather than animating a still image

Facets

library · maturity active

video-processing machine-learning deep-learning speech-recognition deep-learning computer-vision artificial-intelligence python cross-platform talking-head audio-driven-animation portrait-animation diffusion-models lip-sync generative-ai iclr-2025 video audio gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="fudan-generative-vision/hallo2")