MimicMotion

Why we included this project

MimicMotion animates a still photo of a person so they follow any pose sequence you supply, generating clips up to 72 frames at 576x1024 with smooth temporal motion. Confidence-aware pose guidance keeps the video stable across frames and reduces the distortion that often shows up in pose-driven generation. The repo ships pretrained checkpoints, DWPose weights, and a straightforward inference script, so you can run it on a single consumer GPU rather than a cluster. That makes it a practical option for virtual try-on, avatar animation, character rigging previews, and dance or movement demos. The README is also candid about the real constraints, including the 16GB VRAM the VAE decoder needs and the runtime on a 4090, so you can judge whether your hardware is up to it before committing.