# KlingAIResearch/LivePortrait

**Attribution required: if you use, quote, or summarise this content, you must credit and link back to [awesome-repositories.com](https://awesome-repositories.com/repository/klingairesearch-liveportrait).**

17,830 stars · 1,848 forks · Python · other

## Links

- GitHub: https://github.com/KlingAIResearch/LivePortrait
- Homepage: https://liveportrait.github.io
- awesome-repositories: https://awesome-repositories.com/repository/klingairesearch-liveportrait.md

## Topics

`face-animation` `image-animation` `video-editing` `video-generation`

## Description

LivePortrait is a computer vision framework designed for portrait animation and generative video synthesis. It functions as a deep learning system that transfers facial expressions and head movements from a driving video source onto a static image or an existing portrait video, effectively decoupling the subject's identity from the dynamic motion patterns.

The framework utilizes keypoint-based motion retargeting and implicit 3D latent representations to map movements across different subjects, including both human and animal portraits. By employing canonical motion normalization and feature-space non-rigid warping, the system ensures that driving inputs are standardized and aligned with the target geometry before synthesis occurs.

The process includes generative adversarial refinement to maintain visual fidelity and skin texture, alongside mask-based stitching to integrate the animated facial region into the original background. The framework also supports inference optimization through pre-computed motion templates and input scaling to facilitate the production of consistent, high-quality video output.

## Tags

### Artificial Intelligence & ML

- [Facial Animation](https://awesome-repositories.com/f/artificial-intelligence-ml/facial-animation.md) — Maps facial expressions from driving videos onto target portraits while preserving subject identity.
- [AI Video Generators](https://awesome-repositories.com/f/artificial-intelligence-ml/ai-video-generators.md) — Synthesizes realistic facial animations by mapping motion features from source media onto target subjects.
- [Portrait Video Retargeting](https://awesome-repositories.com/f/artificial-intelligence-ml/generative-ai-resources/diffusion-visual-models/generative-ai-tasks/video-to-video-synthesis/portrait-video-retargeting.md) — Applies motion from a driving source to an existing portrait video to modify or retarget facial expressions and head movements. ([source](https://cdn.jsdelivr.net/gh/KlingAIResearch/LivePortrait@main/README.md))
- [Video Generation](https://awesome-repositories.com/f/artificial-intelligence-ml/video-generation.md) — Synthesizes realistic facial animations by mapping motion features from source media onto target subjects.
- [Facial Landmark Detection](https://awesome-repositories.com/f/artificial-intelligence-ml/facial-landmark-detection.md) — Maps facial landmarks from a driving source to a target image to decouple identity from dynamic movement patterns.
- [Generative Adversarial Networks](https://awesome-repositories.com/f/artificial-intelligence-ml/generative-adversarial-networks.md) — Uses a discriminator network to ensure the generated facial features maintain high visual fidelity and realistic skin texture.
- [Non-Rigid Warping Utilities](https://awesome-repositories.com/f/artificial-intelligence-ml/feature-alignment/non-rigid-warping-utilities.md) — Applies deformations to source image features to align them with target motion before final pixel synthesis occurs.
- [Latent Space Generative Models](https://awesome-repositories.com/f/artificial-intelligence-ml/generative-ai-resources/diffusion-visual-models/generative-ai-models/latent-space-generative-models.md) — Encodes facial geometry and expression as latent features to allow for consistent spatial transformations across different portrait subjects.
- [Inference Acceleration](https://awesome-repositories.com/f/artificial-intelligence-ml/inference-acceleration.md) — Optimizes model execution using pre-computed motion templates or just-in-time compilation to reduce the total processing time required for generating high-quality video content. ([source](https://cdn.jsdelivr.net/gh/KlingAIResearch/LivePortrait@main/README.md))
- [Inference Optimization](https://awesome-repositories.com/f/artificial-intelligence-ml/machine-learning/infrastructure/model-inference-serving/inference-optimization.md) — Reduces the processing time required for generating high quality video content through efficient model execution and motion template usage.
- [Coordinate Normalization Utilities](https://awesome-repositories.com/f/artificial-intelligence-ml/coordinate-normalization-utilities.md) — Standardizes driving video inputs into a shared coordinate space to ensure stable and predictable animation across diverse source media.
- [Implicit Latent Representations](https://awesome-repositories.com/f/artificial-intelligence-ml/representation-probing/implicit-latent-representations.md) — Encodes facial geometry and expression as latent features to allow for consistent spatial transformations across different portrait subjects.

### Graphics & Multimedia

- [Portrait Animation Engines](https://awesome-repositories.com/f/graphics-multimedia/media-processing-analysis/face-portrait-manipulation/portrait-animation-engines.md) — Transfers facial expressions and head movements from a driving video or motion template onto a static portrait image. ([source](https://cdn.jsdelivr.net/gh/KlingAIResearch/LivePortrait@main/README.md))
- [Image Blending](https://awesome-repositories.com/f/graphics-multimedia/media-processing-analysis/face-portrait-manipulation/image-blending.md) — Blends the synthesized facial region back into the original background frame using mask-based composition to preserve global consistency.
- [Mask-Based Blending Logic](https://awesome-repositories.com/f/graphics-multimedia/media-processing-analysis/face-portrait-manipulation/image-blending/mask-based-blending-logic.md) — Blends the synthesized facial region back into the original background frame using composition to preserve global visual consistency.
- [Animal Motion Synthesis](https://awesome-repositories.com/f/graphics-multimedia/animation-motion/animal-motion-synthesis.md) — Detect and map keypoints from animal source images to animate them using specialized motion templates and driving data for realistic movement across different species. ([source](https://cdn.jsdelivr.net/gh/KlingAIResearch/LivePortrait@main/README.md))
- [Facial Compositing](https://awesome-repositories.com/f/graphics-multimedia/media-processing-analysis/media-manipulation/media-processing-workflows/video-transformation-enhancement/programmatic-video-production/facial-compositing.md) — Integrates generated facial features into original video frames using mask-based stitching for seamless visual consistency.