---
license: other
base_model: black-forest-labs/FLUX.2-klein-base-4B
tags:
- face-restoration
- blind-face-restoration
- reference-guided-restoration
- flux
- lora
- computer-vision
pipeline_tag: image-to-image
---
# IConFace LoRA Adapter
This repository hosts the released LoRA adapter for **IConFace: Identity-Structure
Asymmetric Conditioning for Unified Reference-Aware Face Restoration**.
- Project page: https://cosmicrealm.github.io/IConFace/
- arXiv: https://arxiv.org/abs/2605.02814
- Code: https://github.com/cosmicrealm/IConFace
- Model weights: https://huggingface.co/zhangjinyang/IConFace
## Visual Results
### Reference-Aware Restoration
### No-Reference Restoration
More visual comparisons are available on the project page.
## Files
```text
iconface_lora.safetensors
```
This adapter is used with:
- Base model: `black-forest-labs/FLUX.2-klein-base-4B`
- LoRA target preset: `flux2_klein`
- LoRA rank: 16
- Expected path in the code release: `pretrained_models/iconface_lora.safetensors`
## Usage
Use the GitHub code release:
```bash
git clone https://github.com/cosmicrealm/IConFace.git
cd IConFace
pip install -r requirements.txt
bash scripts/download_models.sh
```
No-reference inference:
```bash
python scripts/infer_no_ref.py \
--model_dir pretrained_models/FLUX.2-klein-base-4B \
--lora_path pretrained_models/iconface_lora.safetensors \
--adaface_model_path pretrained_models/adaface_ir50_ms1mv2.ckpt \
--condition_image examples/lq.png \
--output_dir outputs/no_ref_demo
```
Reference-aware inference:
```bash
python scripts/infer_with_ref.py \
--model_dir pretrained_models/FLUX.2-klein-base-4B \
--lora_path pretrained_models/iconface_lora.safetensors \
--adaface_model_path pretrained_models/adaface_ir50_ms1mv2.ckpt \
--condition_image examples/lq.png \
--reference_image examples/ref1.png \
--reference_image examples/ref2.png \
--reference_image examples/ref3.png \
--output_dir outputs/ref_demo
```
## Citation
```bibtex
@misc{niu2026iconface,
title={IConFace: Identity-Structure Asymmetric Conditioning for Unified Reference-Aware Face Restoration},
author={Niu, Axi and Zhang, Jinyang and Qing, Senyan},
year={2026},
eprint={2605.02814},
archivePrefix={arXiv},
primaryClass={cs.CV},
doi={10.48550/arXiv.2605.02814},
url={https://arxiv.org/abs/2605.02814}
}
```
## License Notes
This adapter depends on the FLUX.2-Klein-base-4B base model. Users must comply
with the license and access terms of the base model and all external identity
models used by the code release.