explaingit

klingairesearch/liveportrait

★ 18,370PythonAudience · developerComplexity · 3/5MaintainedSetup · hard

TLDR

AI system that animates still portrait photos to move naturally, driven by a video or image reference. Works on faces and animals with local or browser-based control.

Mindmap

mindmap
  root((repo))
    What it does
      Animate portraits
      Drive from video
      Drive from image
    Key features
      Stitching control
      Retargeting motion
      Animal support
    How to use
      Browser interface
      Local installation
      Windows one-click
    Tech stack
      Python
      Neural networks
      Deep learning
    Platforms
      Linux Windows macOS
      Kuaishou Douyin WeChat

Things people build with this

USE CASE 1

Animate a still photo of a person to match head movements and expressions from a video.

USE CASE 2

Create lip-synced video of a portrait by driving it with audio or a reference video.

USE CASE 3

Animate pet photos (cats, dogs) to move naturally using a motion reference.

USE CASE 4

Build a video editing tool that lets users control portrait animation interactively in a browser.

Tech stack

PythonPyTorchDeep LearningComputer Vision

Getting it running

Difficulty · hard Time to first run · 1h+

Requires PyTorch installation, pre-trained model downloads, and GPU/CUDA for reasonable performance; CPU-only is very slow.

License could not be detected automatically. Check the repository's LICENSE file before use.

In plain English

LivePortrait is a Python research project for animating portrait images using AI. Given a still photo of a face, the system can make it move in a lifelike way by using a driving video or image as a reference for the motion, for example, making a portrait smile, turn its head, or lip-sync to audio. The project comes from researchers at Kuaishou Technology and has been adopted in video platforms including Kuaishou, Douyin, and WeChat Channels. It supports animating both human faces and animals (such as cats and dogs). The system includes controls for stitching (blending the animated result back with the original image naturally) and retargeting (mapping motion from a driver source onto the portrait). It can be run locally on Linux, Windows, or macOS with Apple Silicon, and the repository includes a browser-based interface for interactive use. A one-click installer for Windows is available. Modes include driving from a video, driving from an image, and editing a portrait video. A hosted online demo is also available. The code and associated neural network models are provided for research and inference use.

Copy-paste prompts

Prompt 1
How do I set up LivePortrait locally on my Mac with Apple Silicon to animate portrait images?
Prompt 2
Show me how to use the browser interface to drive a portrait animation from a video file.
Prompt 3
What's the difference between stitching and retargeting in LivePortrait, and when would I use each?
Prompt 4
Can I animate a photo of my dog using LivePortrait, and what format should the driving video be in?
Prompt 5
Walk me through the Windows one-click installer for LivePortrait and how to run the first animation.
Open on GitHub → Explain another repo

Generated 2026-05-18 · Model: sonnet-4-6 · Verify against the repo before relying on details.