Clean ligne-claire graphic illustration look: fine ink outlines, flat muted colour, calm everyday scenes. A STYLE LoRA for MiniMax H3 video - it changes the look, not the subject.

EXPERIMENTAL: lightly tested - the two clips on this page are the main test so far. This is a drawn style and H3's base pulls toward photorealism, so it runs a little above 1.0: the samples use 1.1. If the look doesn't take on your prompt, turn it up (1.2 to 1.5) and tell me what you find.

Trigger word: kreasammelinstyle (put it first in the prompt). Strength 1.1 (1.0 to 1.5). One file: epoch 50.

Base: MiniMax H3 (fl2va). Loads in ComfyUI with LoraLoaderModelOnly (the file must sit in models/loras/), or in WanGP under loras/minimax_h3.

How it was made: Krea 2 dataset generated with the source style LoRA, then hand-curated down to the 89 best images (every subject unique, about half people, cinematic shot vocabulary in every caption); trained with Fizgig "MiniMax H3 Style (LoRA 8)" - rank 8, LR 2e-4, 50 epochs, blocks 0-3,6-47, adamw, shift 0.6667.

Samples: two text-to-video clips, rendered locally in WanGP with MiniMax H3 FL2VA (pruned 20B) at 1920x1080 "Two Phases" (half-res pass -> latent upscale -> refine), 4 steps, euler, seed 77, 5 s; a 4-step distill LoRA (larryvrh v4) on phase 1 and this LoRA at 1.1 on both phases. Plain text prompts with no reference images, so the look you see is the LoRA. Audio stripped.

Credit: derived from "KreaSammelin" by Adel_AI (Krea 2) - https://civarchive.com/models/2899227/kreasammelin?modelVersionId=3278046 - please support the original creator; it is also linked under Recommended Resources.