csbowei/ART

[ECCV 2026] Anchoring on Reality: Breaking the Pseudo-Target Ceiling in Makeup Transfer

Bo Wei ^1*  Xianhui Lin ^2†  Yi Dong ²  Zhongzhong Li ²  Zonghui Li ²  Zirui Wang ² 

Jiachen Yang ²  Xing Liu ²  Hong Gu ²  Xiaoming Li ^3✉  Wangmeng Zuo ^1✉

¹ Harbin Institute of Technology  ² vivo BlueImage Lab  ³ Nanjing University

^* Work done during an internship at vivo.  ^† Project lead

   

This repository provides the model weights for ART: Anchoring on Reality: Breaking the Pseudo-Target Ceiling in Makeup Transfer.

ART is a makeup transfer framework that applies a reference cosmetic style to a source face while preserving identity, geometry, and fine-grained makeup details.

Files

File	Task
art_transfer_lora_512.safetensors	Makeup transfer
art_transfer_lora_1024.safetensors	Makeup transfer
art_demakeup_lora_512.safetensors	Makeup removal
art_demakeup_lora_1024.safetensors	Makeup removal

Higher-resolution checkpoints will be released in the future. Stay tuned.

Usage

These LoRA adapters are for black-forest-labs/FLUX.1-Kontext-dev.

Refer to the project repository: https://github.com/csbowei/ART

Example:

# Makeup transfer: apply the reference's makeup onto the input face
python infer.py \
    --task        mt \
    --input       ./examples/source/src_001.jpg \
    --ref         ./examples/ref/ref_001.jpg \
    --lora_path   /path/to/art_transfer_lora_1024.safetensors \
    --resolution  1024

# Makeup removal: strip cosmetics from the input face
python infer.py \
    --task        demakeup \
    --input       ./examples/ref/ref_002.jpg \
    --lora_path   /path/to/art_demakeup_lora_1024.safetensors \
    --resolution  1024

Citation

@article{wei2026art,
  title={Anchoring on Reality: Breaking the Pseudo-Target Ceiling in Makeup Transfer},
  author={Wei, Bo and Lin, Xianhui and Dong, Yi and Li, Zhongzhong and Li, Zonghui and Wang, Zirui and Yang, Jiachen and Liu, Xing and Gu, Hong and Li, Xiaoming and Zuo, Wangmeng},
  journal={arXiv preprint arXiv:2606.31089},
  year={2026}
}