LivePortrait Mirror (A.I.M.I)

Mirror of KwaiVGI/LivePortrait plus its required InsightFace buffalo_l face-detection bundle, re-hosted for stable URLs inside A.I.M.I's Animate-tab Portrait mode. Contents unmodified.

LivePortrait takes a single portrait photo + a driving video and animates the photo with the video's expressions, head motion and lip movements. Used by A.I.M.I as the "Portrait" mode in the Animate tab.

Files

liveportrait/base_models/ β€” core animation

  • appearance_feature_extractor.pth (~3 MB)
  • motion_extractor.pth (~107 MB)
  • warping_module.pth (~174 MB)
  • spade_generator.pth (~212 MB)

liveportrait/retargeting_models/

  • stitching_retargeting_module.pth (~2 MB)

liveportrait/

  • landmark.onnx (~109 MB) β€” face landmarks

insightface/models/buffalo_l/ β€” face detection (required by LivePortrait pipeline)

  • det_10g.onnx (~16 MB)
  • 2d106det.onnx (~5 MB)

Total: ~630 MB.

License

MIT (LivePortrait code + weights). InsightFace buffalo_l weights also MIT. Redistributed unchanged.

Attribution

  • LivePortrait: Jianzhu Guo, Dingyun Zhang, Xiaoqiang Liu, Zhizhou Zhong, Yuan Zhang, Pengfei Wan, Di Zhang β€” LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control (Kuaishou Visual Generation and Interaction Center β€” KwaiVGI, 2024).
  • InsightFace buffalo_l: Jia Guo, Jiankang Deng et al. β€” InsightFace project.
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support