YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Also on Civitai.
license: other license_name: non-commercial license_link: LICENSE.md language: - en tags: - lora - text-to-image - image-generation - krea2 - style - style-transfer - painterly - impasto - thick-paint - comfyui
magful — Krea 2 Thick-Paint Style LoRA
A thick-paint / impasto style LoRA for Krea 2 with a built-in colour strategy — not just a texture filter. It doesn't lock composition and doesn't override your art style.
File: krea2_style_magful.safetensors (109 MB, rank 16)
Base model: Krea 2
Trigger word: magful — optional, the LoRA is always-on
Quick start
| Item | Value |
|---|---|
| Recommended strength | 1.0 |
| Recommended cfg | 1.0 (daily) / 1.5 (punchier) |
| Sampler | euler / sgm_uniform |
| Steps | 8 |
| CLIP | Qwen3-VL 4B |
| VAE | qwen_image_vae |
Clone / download:
pip install -U huggingface_hub
huggingface-cli download weiermod/magful-krea2-lora krea2_style_magful.safetensors \
--local-dir ComfyUI/models/loras
Direct link:
https://huggingface.co/weiermod/magful-krea2-lora/resolve/main/krea2_style_magful.safetensors
What it does
Controlled experiment (fixed prompt × 4 seeds × 3 strengths, 36 renders):
LoRA colorfulness = 0.896 × base colorfulness + 3.93 R² = 0.995 (strength 1.0)
Crossover at base ≈ 37.7 — below it the LoRA adds colour, above it mutes.
| Fixed prompt | Base | Strength 1.0 | Ratio |
|---|---|---|---|
| Gothic / dark (nun in cathedral) | 15.1 | 18.5 | 1.23 ↑ |
| Mid-tone (afternoon street) | 35.8 | 34.5 | 0.96 ≈ |
| Saturated (magenta/cyan flower field) | 95.8 | 69.9 | 0.73 ↓ |
Both directions: distinguishable colours +40% ~ +174%, luminance micro-detail +3% ~ +49%, stronger colour peaks (the most saturated 5%), especially in dark scenes.
It doesn't simply "add saturation" or "go grey" — it pulls every starting point toward a "composed but powerful" middle range.
⚠️ Want an empty scene / scenery only?
It can do empty scenes — short prompts just can't hold back its default configuration.
38 empty-scene renders (6 scene types × 3 seeds):
| Prompt style | Characters appeared |
|---|---|
| Short (~10 words) + poster-art prefix | 18 / 18 |
Short + no humans at the very front |
14 / 14 (control included) |
| Medium detail (~47 words) | 0 / 2 |
| Long prompt (200+ words, per-object spec) | 0 / 2 |
| Still-life subject (no poster prefix) | 0 / 6 |
dark fantasy aesthetic tone |
0 / 4 |
The official poster art / cinematic composition prefix is the main switch — it turns even a
still life into a portrait, and moving no humans to the front does not fix it.
Empty scenes come from writing detail, not from adding negatives.
Controllability (all A/B tested)
| Item | Result |
|---|---|
| Explicit hair colour | 13/13 (black / blonde / green / pink / brown / silver / purple) |
| Explicit head-count on plain background | 6/6 exact |
| Explicit photographic style | 8/8 photoreal — LoRA steps aside |
| Six forced media (ink / oil / 3D / photo / watercolour / vector) | 6/6 kept their medium — not invasive |
With ink line art forced, luminance gradient changed by 0.0002 (0.1152 vs 0.1154). Prompt responsiveness: within-group distance 0.31 (LoRA) vs 0.36 (base) — distinguishing ability preserved, only ~13% homogenisation.
Things to watch
- Unspecified hair → tends to silver (the base model already prefers this)
- Unspecified outfit → tends to school uniform (6/6 vs 6/6 casual for base)
- Scenes implying company → may add a second figure; add
solo, only one person in frame - Short prompt + poster prefix → fills the frame with people
Retracted claims
Several earlier conclusions were overturned by later controlled tests — including three from this project's own earlier drafts. Listed so nobody propagates them:
| Once claimed | Actually |
|---|---|
| "Overrides explicit photographic style" | 8/8 stayed photoreal |
| "Miscounts people in group scenes" | 6/6 exact with 3girls on a plain background |
| "Obeys empty-scene prompts 4/4" | that prompt read solo, 1girl, only one person in frame — it asked for a person |
| "Has a colorfulness floor around 28–31" | dark scenes output 15.6–20.8 — no floor; the real relation is linear shrinkage |
| "+91% colorfulness in dark scenes" | uncontrolled single sample; controlled value +10~20% |
| "Lighting words are bound to characters" | the no-word control also produced people 14/14; prompt specificity is the deciding factor |
"Silver hair" is not this LoRA's signature — the base model draws silver-haired girls from
1girl alone.
Example prompts
Reusable prefix:
magful, Modern high-end anime illustration, official poster art, cinematic composition, intricate details, sophisticated style, depth of field, sharp focus on subject, masterpiece.
1girl, gothic nun, silver cross, gothic cathedral, heavy rain, dark dramatic lighting1girl, valkyrie, silver winged helmet, ornate plate armor, holding a spear, divine golden light, clouds and sky, epic battlefield, dramatic backlighting1girl, wandering swordswoman, black hair in a high ponytail, worn traveler cloak, katana at her side, cherry blossom grove, wind scattering petals, twilight1girl, young librarian, brown hair with glasses, cream blouse, standing among towering bookshelves, warm lamplight, dust motes, cozy atmosphere1girl, elf archer, long green hair, pointed ears, dark green cloak with gold embroidery, drawing a wooden bow, ancient mossy forest, fireflies, moonlight1girl, cosmic priestess, white hair, layered white and gold robe, standing before a giant celestial harp, starfield, floating planets, ornate golden architecture
Negative:
cartoon, chibi, cel shading, flat colors, childlike, photorealistic, 3d render, low quality, bad anatomy, text, watermark
Training
- Base: Krea 2 ｜ trainer: ai-toolkit ｜ 2500 steps / 138 epochs ｜ rank 32
- Dataset: self-curated (mostly own AI-generated works, plus a small number of third-party AI-generated pieces), style-consistent but structurally diverse; not retained after training.
License
Non-commercial. Because the dataset includes third-party material, do not use commercially. Derivatives should credit the source and keep the same restriction. See LICENSE.md.
中文说明
magful —— Krea 2 厚涂风格 LoRA。 触发词 magful 非必须，默认生效。推荐强度 1.0，cfg 1.0（日常）/ 1.5（更艳）。
想做空景？ 关键在提示词的具体程度，不在题材。38 张空景测试：短句（10 词）+ 海报前缀 → 18/18 出人；加到 47 词 → 0/2；234 词 → 0/2；静物题材 → 0/6；dark fantasy aesthetic 调性 → 0/4。official poster art 这类前缀是主要开关，把 no humans 放最前面救不回来。空景靠「写详细」，不靠「加负面」。
效果：LoRA 鲜艳度 = 0.896 × 基座鲜艳度 + 3.93（R²=0.995），交叉点约 37.7 —— 低于它加色（暗调 ×1.23），高于它减色（高饱和 ×0.73）。两个方向都提升色阶（+40174%）和明暗微细节（+349%）。
可控性：发色 13/13、白底人数 6/6、摄影风格 8/8 让位、六种强制画风 6/6 全部守住（墨线风格下梯度只差 0.0002）。
注意：不写发色会银白、不写服装会制服、场景暗示有同伴会多一个人。
许可：非商用。素材含第三方 AI 作品，训练后未保留。