ICLR 2026

Durian
Dual Reference Image-Guided Portrait
Animation with Attribute Transfer

Hyunsoo Cha Byungjun Kim Hanbyul Joo

Seoul National University

Paper arXiv OpenReview Code (Coming Soon)

TL;DR — Given a portrait image and single or multiple reference images specifying target attributes (e.g., hairstyle, eyeglasses), Durian generates a portrait animation with facial attribute transfer conditioned on a keypoint sequence.

◆
Demonstrations

Results

All result videos are complete zero-shot inference results not included in the training set, without any additional training or optimization. We present results of transferring attributes to the portrait while simultaneously animating it.

Hair Transfer

Hat Transfer

Beard Transfer

Eyeglasses Transfer

◆
Extended Capabilities

Applications

All result videos are complete zero-shot inferences, showing the transfer of multiple attributes simultaneously to the portrait or interpolation between attributes.

Multi Attribute Transfer

Attribute Interpolation

Citation

BibTeX

@inproceedings{cha2026durian,
  author={Cha, Hyunsoo and Kim, Byungjun and Joo, Hanbyul},
  booktitle={International Conference on Learning Representations},
  title={Durian: Dual Reference Image-Guided Portrait Animation with Attribute Transfer},
  year={2026}
}