A graphic novel style lora. I’m going for digital art this time.
SagaRealm — Klein Issue
SagaRealm is a LoRA for creating graphic novel style images. Trained on Klein 9B Base, works on both Klein 9B and Klein 9B Base.
One LoRA, four distinct styles — all triggered by how you open your prompt:
SagaRealm style.
The default. Bold graphic novel aesthetic, confident linework, saturated palette. Use this when you want the style to do the talking.
A graphic ink illustration in SagaRealm style.
High-contrast ink linework, clean contours, solid colour blocking. Your comic panel, your book cover.
Painterly chiaroscuro in SagaRealm style.
No linework. Form sculpted entirely through light and shadow. Put your subject somewhere dark — the style needs darkness to work against.
An ethereal, abstract composition in SagaRealm style.
Flowy lines, vaporous textures, spectral gradients. Weightless and dreamlike.
Always start your prompt with the trigger phrase. That’s what tells the model which style to use before anything else.
I use a LoRA weight between 0.8 and 1.0. Start at 0.9 and adjust from there.
Works well for characters, creatures, environments and scenes. Handles both male and female subjects. Likeness transfer from a reference image works cleanly.
FLUX and Z-Image Turbo:
The trigger is SagaRealm style but I built in three styles. You can choose any of the triggers I defined.
SagaRealm style. chiaroscuro painting.
SagaRealm style. A graphic ink illustration.
SagaRealm style. An ethereal, abstract composition.
Description
Much more editing on the dataset, especially the captioning strategy.
FAQ
Comments (5)
Z-Image Turbo picks up art styles so well. Do you know if Z-Image Base does too?
I’ve read that they aren’t really compatible with each other, base and turbo. It’s on my radar though.
@maw6011 Ah, I've heard it trains concepts much better into Turbo, and the compatibility is more about strength. Loras trained in Base require about twice as much strength in Turbo, but the subject does come through, and with better adaptability to different situations/compositions/loras. At least, that's the info I got from several people who train loras.
I haven't heard much about training styles though. That could be completely different.
@Jellai I’m doing a test training right now.
This is fantastic work
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.