MagicPose
Realistic human pose and facial expression transfer from video.
About
MagicPose is an identity-aware diffusion model for retargeting human poses and facial expressions from a driving video onto a still reference image. It preserves the subject's appearance, clothing, and identity while producing video with new motion and expressions, and is compared in the paper against MagicAnimate, Disco, and TPS. Code, pretrained models, and a Hugging Face demo are released by the USC and ByteDance authors.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Category
- AI Animation & Motion
- Price
- Free
- Platform
- Local/Desktop
- Difficulty
- Advanced (4/5)
- Minimum VRAM
- 12 GB
- Added
- Apr 3, 2026
Related Tools
Animates a still human photo with 3D SMPL parametric motion guidance extracted from a driving video.
Free markerless motion capture system that works with ordinary cameras and no special hardware.
Audio-driven Tencent model that animates avatar images into emotion-controllable dialogue videos.
ByteDance's audio-conditioned latent diffusion model for lip-syncing video to new speech.
Audio-driven talking head animation from a single image.
Effective whole-body pose estimation with few-shot keypoint detection.