Give your story a painted cinematic world
Sculptural forms. Painted colour planes. Light that tells a story. ARKANE is my Krea 2 style LoRA inspired by the visual language of Arcane.
A glance across a workshop. An unexpected encounter with a dragon. A city waiting for dawn. I built this style to carry that sense of presence into your own characters and worlds, with distinct materials, selective detail and quiet surfaces between the accents.
Trigger: silbrioarkane
Start at LoRA strength 1.0.
Set the scene
Describe your character or subject and their gesture.
Choose the camera position and the light that best tell the story.
Give the image a clear emotional moment: an encounter, a decision, a quiet pause. Your own characters and worlds are welcome.
The gallery recipe
Krea 2 Turbo · 8 steps · CFG 1
Euler / Simple · LoRA strength 1.0
Trained on Krea 2 RAW with 60 curated images.
Character scenes are the heart of this release. Elaborate interactions and specialised objects may need several attempts; inspect hands, props and spatial connections in the result.
What story will you tell?
If ARKANE sparks an idea, share your images in the gallery with the LoRA attached. Follow my profile to discover the next experiments, and let me know what you enjoyed creating.
This is an independent fan-made style study, unaffiliated with any studio or artist.
Description
ARKANE · MOSS Image
A cinematic style LoRA inspired by the painted visual language of Arcane. Sculptural volumes, expressive faces, broad painted color planes and quiet surfaces carry the look across characters, landscapes, architecture and still lifes.
Trigger: silbrioarkane · LoRA strength: 1.0.
Gallery settings: MOSS-Image-Base BF16 · 25 steps · CFG 2.5 · ER-SDE / Simple · denoise 1.0. Qwen3.5 9B INT8 ConvRot text encoder and FLUX.2 VAE. Prompt enhancement off. New images rendered at 1344 × 2016, 2016 × 1344 or 1664², then finished to a 4096 px long edge with IllustrationJaNai DAT2 and Lanczos. The reserved comparison examples use their original native dimensions.
Training: 60 curated images · 1,800 steps · rank/alpha 32.
Free download. Standalone adapter trained by silbrio for OpenMOSS MOSS-Image-Base. The upstream base is distributed under Apache License 2.0.









