Downloads · 30 days
674
20% of all-time downloads
thedeoxen/refcontrol-FLUX.2-klein-9B-reference-normal-lora
refcontrol-FLUX.2-klein-9B-reference-normal-lora is a image-to-image model from thedeoxen. Use it when you need one image transformed into another. It is set up for diffusers. The card lists the license as apache-2.0.
A LoRA for FLUX.2 Klein 9B Base that fuses a reference image (identity) with a normal map (structure / surface orientation). It preserves identity and style from the reference while following the 3D shape and composit…
Downloads · 30 days
674
20% of all-time downloads
All-time downloads
3.3K
Public
Repo size
182 MB
Likes
17
Public
Click a slice to open those files.
.safetensors166 MB · 91%
From the Hugging Face model README
A LoRA for FLUX.2 Klein 9B Base that fuses a reference image (identity) with a normal map (structure / surface orientation).
It preserves identity and style from the reference while following the 3D shape and composition from the normal control map.
Trigger word: refcontrol
Each preview is a single combined image from ComfyUI: Normal → Reference → Result (left to right).
<table> <thead> <tr> <th>Normal → Reference → Result</th> </tr> </thead> <tbody> <tr> <td><img src="https://huggingface.co/thedeoxen/refcontrol-FLUX.2-klein-9B-reference-normal-lora/resolve/main/images/a (1).png" width="960" /></td> </tr> <tr> <td><img src="https://huggingface.co/thedeoxen/refcontrol-FLUX.2-klein-9B-reference-normal-lora/resolve/main/images/a (2).png" width="960" /></td> </tr> <tr> <td><img src="https://huggingface.co/thedeoxen/refcontrol-FLUX.2-klein-9B-reference-normal-lora/resolve/main/images/a (3).png" width="960" /></td> </tr> <tr> <td><img src="https://huggingface.co/thedeoxen/refcontrol-FLUX.2-klein-9B-reference-normal-lora/resolve/main/images/a (4).png" width="960" /></td> </tr> <tr> <td><img src="https://huggingface.co/thedeoxen/refcontrol-FLUX.2-klein-9B-reference-normal-lora/resolve/main/images/a (5).png" width="960" /></td> </tr> </tbody> </table> ---This LoRA was primarily trained on humans, but it also works with stylized characters and some objects.
Its main purpose is to preserve identity — facial features, hairstyle, clothing, or object details — from the reference image, while adapting the subject to the structure and surface orientation defined by the normal map.
FLUX.2 Klein 9B Base already handles reference + normal-guided transfer reasonably well with the right prompt alone. This LoRA builds on that capability — it improves consistency and better preserves character identity and normal-map fidelity than the base model without LoRA.
Part of the RefControl family: reference + control fusion for consistent, controllable generation on FLUX.2 Klein 9B Base.
refcontrol in your prompt.Normal map extraction in the included workflow uses DSINE-NormalMapPreprocessor (via comfyui_controlnet_aux):
https://github.com/Fannovel16/comfyui_controlnet_aux
You can disable the built-in normal preprocessor if your input is already a normal map image.
Trained on and recommended with black-forest-labs/FLUX.2-klein-base-9B.
The undistilled Base variant is intended for LoRA training and custom pipelines (~50 inference steps, guidance_scale ~4.0).
The LoRA also works with the 4-step distilled black-forest-labs/FLUX.2-klein-9B for faster inference (~4 steps, guidance_scale ~1.0), but quality may be slightly lower — especially for identity and normal-map fidelity — compared to the Base model.
refcontrol
flux2_klein_9b_refcontrol_normal.safetensorsrefcontrol_normal_flux_klein_9b.json