Accepted paper ACM SIGGRAPH 2026

HairPort

In-context 3D-Aware Hair Import and Transfer for Images

A. Heidari, A. Alimohammadi, W. Michel Pinto Lira, A. Bar-Lev, and A. Mahdavi-Amiri

HairPort transfers a reference hairstyle onto a source face with 3D-aware pose alignment, realistic bald conversion, and identity-preserving synthesis.

SIGGRAPH 2026 paper Code preview Baldy dataset LoRA weights
Code Dataset Weights Citation Paper coming soon Video coming soon

Paper

HairPort: In-context 3D-Aware Hair Import and Transfer for Images

ACM SIGGRAPH 2026 Accepted Paper

HairPort separates bald source generation from hairstyle transfer, aligns reference hair through a 3D-aware stage, and synthesizes the final result from source-aligned hair evidence.

Paper coming soon Video coming soon BibTeX

Results

Robust transfers across identity, style, and viewpoint.

HairPort handles realistic photos, stylized references, large pose gaps, and significant scale differences while preserving the source identity.

HairPort teaser showing multiple source faces, reference hairstyles, and transferred results.
Figure 1. Hairstyle transfer examples across challenging identity, pose, scale, and style differences.

Highlights

A 3D-aware route through hard hair-transfer cases.

3D-aware alignment

Reference hair is reconstructed and re-rendered from the source viewpoint before synthesis.

Large pose gaps

Designed for cases where 2D warping or local editing loses hair structure and scale.

Identity preservation

Bald source generation and hairstyle import are separated to keep the face stable.

Bald Converter + Baldy

FLUX LoRA-based bald conversion trained with 6,400 paired bald/original images.

Method

A pipeline built around source-aligned hairstyle evidence.

The method removes source hair, aligns reference hair with 3D geometry, and conditions final synthesis on the bald source, aligned hair rendering, reference image, and text prompt.

HairPort method diagram showing bald conversion, 3D-aware hair transfer, and final synthesis.
Figure 2. HairPort method overview.

Bald Converter and Baldy

A dedicated bald-generation stage for cleaner transfer.

The Bald Converter produces realistic bald versions of source portraits using LoRA-based in-context adaptation of FLUX. Baldy provides paired bald/original supervision across diverse identities and conditions.

Baldy dataset LoRA weights
Baldy dataset generation and Bald Converter finetuning diagram.
Figure 3. Baldy dataset generation and Bald Converter finetuning.

Code and Assets

Source preview, dataset, and LoRA weights are available.

The implementation snapshot is available now. Packaging, dependency manifests, and installed console-script entry points are still being finalized.

GitHub HairPort source code Official SIGGRAPH 2026 implementation snapshot. README Installation notes Current setup, source preview caveats, and usage examples. Dataset Baldy 6,400 paired bald/original images. Weights Bald Converter LoRAs Adapters for in-context bald generation.

News

Latest updates.

  1. HairPort was accepted to ACM SIGGRAPH 2026.
  2. Project page launched at deepmancer.github.io/HairPort.
  3. Initial source preview released; packaging and dependency manifests will be finalized soon.
  4. Baldy dataset and Bald Converter LoRA weights released.

Citation

Use HairPort in your research?

Please cite the ACM SIGGRAPH 2026 paper.

@inproceedings{heidari2026hairport,
  title     = {HairPort: In-context 3D-Aware Hair Import and Transfer for Images},
  author    = {A. Heidari and A. Alimohammadi and W. Michel Pinto Lira and A. Bar-Lev and A. Mahdavi-Amiri},
  booktitle = {ACM SIGGRAPH 2026},
  year      = {2026}
}