
Trained on a manually curated selection of old-school / dark / grotesque fantasy illustrations, this model will give your images a classic fantasy look, with weathered, gritty textures, and distinctive earthy tones.
Read me first for usage tips!
Trigger word
secret_grotesque_fantasyStyle boosters
Recommended
gritty texture, painterly, illustration style, muted tonesOptional
visible paper grain, visible brushstrokes, weathered, intricate linework, organic detailing, earthy color palette, muted moss greens, rust browns, deep sepia shadows, high contrastInsert the style boosters as is, or better yet, request your favourite LLM to integrate them into the prompt.
Alternatively, describe photo-realistic images for more subtle, yet still very cool results!
Z-Image Base
LoRa strength from 0.4 to 1.0. I like it best at 0.8.
euler as sampler
Low shift: 2-3
CFG 7
trigger word not needed, but style boosters still recommended.
Z-Image Turbo
Use light LoRA weights! I personally prefer to set it between 0.4 and 0.6. Set it to 0.8 to full-blown grotesque fantasy with muted tones, and to 0.2 for just a light touch! A weight of 1.0 is discouraged, but you do you.
When going for the intended painterly illustration style, use euler as sampler, it produces much prettier images. Increasing the shift might help too.
If deformed monsters start to pop up all over your images, take your pills and/or lower the LoRA weight!
Flux
High weights: 0.8 - 1.0.
Try playing with the Flux Guidance to modulate the LoRA effect. If anatomy gets weird, lower the weight and/or increase the Flux Guidance value.
Illustrious
trigger word optional
LoRA weights: 0.7 - 0.9.
Please link your pics and tell me what you think!
Special thanks to @RalFinger for his amazing Lora Captioner and to @ostris for the fantastic AI-toolkit!
Take care and have fun ❤️
SGtW

Description
First final version for Flux
FAQ
Comments (10)
Flux Base please :)
Of course, that's always fine. This style is not meant to be consensual.
@SecretGoTheWolves I conceptually dont know how style can be consensual by not understanding who is giving consent to whom in this situation, but I hope we will get there
@Ses_AI consensual as in relating to consensus, not to consent, in this case
One of the high-quality models (Flux) for creating portraits. It adds additional hyperrealism (in my case) and behaves appropriately when combined with other Loras. Thank you! :)
Thank you for your review :)
So happy to see the little tool I made helped you on your dataset. Also thanks for the purchase of my badge. Keep up the good work! ❤️
Your captioner is well done and very useful; I'll continue to use it for my next datasets.
The gesture of donating me 300 Yellow Buzzers is kind to me. I'll return the sum since I now support the community. I'll add your model to my favorites. https://civitai.com/images/120699613
well, thanks for that. I am not exactly in need of buzz either, since I'm the lucky bastard who won the last contest jackpot - I'll pass on the 300... and then some, to the next guy... ;)
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.












