HarmoVid: Relightful Video Portrait Harmonization

Jun Myeong Choi1, Jae Shin Yoon2, Luchao Qi1, Roni Sengupta1, Joon-Young Lee2
1University of North Carolina at Chapel Hill, 2Adobe Research

CVPR 2026

Paper

Demo Video.

Qualitative comparison of our HarmoVid to other image- or video-based video harmonization methods

Deflickering Comparison on per-frame harmonization output

Ablation study on three-stage pipeline

Ablation study on pseudo-α mask

Result on non-portrait foreground objects

BibTeX

@inproceedings{choi2026relightful,
  title={Relightful Video Portrait Harmonization},
  author={Choi, Jun Myeong and Yoon, Jae Shin and Qi, Luchao and Sengupta, Roni and Lee, Joon-Young},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={23356--23366},
  year={2026}
}