AnyAct Checkpoints

This model repository contains the checkpoint files, source videos, and released data files required by the AnyAct code release.

It includes:

  • AnyAct release checkpoints:
    • checkpoint_dir/anyact/generator/anyact_stage1_3d_traj/model/net_best_acc.tar
    • checkpoint_dir/anyact/generator/anyact_3ds2d_traj/model/latest.tar
  • Original SnapMoGen/MoMask++ pretrained assets required by AnyAct:
    • VQ-VAE under checkpoint_dir/snapmogen/vq/
    • base masked transformer under checkpoint_dir/snapmogen/momask_plus/momaskplus_hrvq3_nlayer8_cdp0.1_ca_bm/
    • evaluator under checkpoint_dir/snapmogen/evaluator/eval_klde-5_late-5_nlayer6_norm/
    • optional GMR under checkpoint_dir/snapmogen/gmr/gmr_d292/
  • Released source videos for VLM evaluation:
    • cropped source videos under data/source_videos/
  • Released caption data:
    • data/all_caption_clean_with_qwen3_8b.json

Usage

Clone or download this release package into the AnyAct project root so that the checkpoint_dir/ and data/ folders appear directly under AnyAct/:

cd AnyAct
git clone https://huggingface.co/LiuhanChen/AnyAct tmp_anyact_release
cp -r tmp_anyact_release/checkpoint_dir ./checkpoint_dir
cp -r tmp_anyact_release/data/source_videos ./data/source_videos
cp tmp_anyact_release/data/all_caption_clean_with_qwen3_8b.json ./data/all_caption_clean_with_qwen3_8b.json

After placement, the AnyAct code should resolve paths such as:

checkpoint_dir/snapmogen/vq/hrvq_nq2_296_nb2048_fk0_gl1.5_ex2_ema_attn/model/net_best_mpjpe.tar
checkpoint_dir/snapmogen/momask_plus/momaskplus_hrvq3_nlayer8_cdp0.1_ca_bm/model/net_best_fid.tar
checkpoint_dir/anyact/generator/anyact_stage1_3d_traj/model/net_best_acc.tar
checkpoint_dir/anyact/generator/anyact_3ds2d_traj/model/latest.tar
data/source_videos/{sample_name}_crop.mp4
data/all_caption_clean_with_qwen3_8b.json

Notes

The snapmogen/ subtree preserves original SnapMoGen/MoMask++ checkpoint names for compatibility. The anyact/ subtree contains the AnyAct release checkpoints used by the AnyAct code.

Please cite AnyAct and SnapMoGen/MoMask++ when using these checkpoints.

Citation

If you use AnyAct checkpoints, data, or code, please cite:

@article{chen2026anyact,
  title={AnyAct: Towards Human Reenactment of Character Motion From Video},
  author={Chen, Liuhan and Zhong, Lei and Wang, Jiewei and Shuai, Qin and Yuan, Li and Fan, Leidong and Li, Qing and Liu, Kanglin},
  journal={arXiv preprint arXiv:2605.15497},
  year={2026}
}
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. 🙋 Ask for provider support

Paper for LiuhanChen/AnyAct