RefControl Flux Kontext β Reference Lineart LoRA
π Short description
A LoRA for Flux Kontext Dev that fuses a reference image (left) with a lineart map (right).
It preserves identity and style from the reference while following the lines and contours from the control.
Trigger word: refcontrollineart
π Extended description
This LoRA was primarily trained on human characters and objects, but it works well with stylized art too.
The main goal is to preserve identity β facial features, clothing, or style β from the reference image, while adapting them to the structure given by the lineart map.
βοΈ How to use
Concatenate two images side by side:
Left: reference image (person, character, or object).
Right: lineart map (black & white contour drawing).
Add the trigger word
refcontrollineartin your prompt.Adjust LoRA weight (recommended 0.7β1.0).
β Example prompt
refcontrolcanny, change ineart to photo with reference from left
π― What it does
Preserves identity and style of the reference.
Uses lineart to control outlines, proportions, and composition.
Works for both realistic and stylized generations.
β‘ Tips
Best results when the lineart map is clear and clean (minimal shading).
π Use cases
Converting sketches to full artworks while keeping reference identity.
Character redraws in different art styles.
Consistent character design across multiple illustrations.
Turning lineart drafts into photorealistic or stylized renders.
Description
FAQ
Comments (5)
thanks)) it works with nunchaku
but on the finale we get two pictures in one. Do you plan to improve Lora so that on the finale there is only a picture with the result?
hi, thank you!
I am planning to change workflow to change workflow to get final image.
Change lora itself it is a bit tricky, because it easier to train when input and output have same size. I'll let you know when I updated workflow.
add image crop node in comfy
i was just thinking that i need something exactly like this. thank you.
Thank you so much!!!





