Classic tabletop-rulebook fantasy illustration style for Krea 2. Dramatic character-first compositions, painted armor and scale texture, torchlit dungeons, golden-hour battlefields — the look of a modern D&amp;D sourcebook plate. Style only: no characters baked in, so your subjects stay yours.USAGE — Trigger: dndstyle (put it at the start of your prompt). Strength 1.0 default; forgiving from 0.5 (light fantasy grade) to 2.0 (full homage). Model-only LoRA: use LoraLoaderModelOnly (there is no CLIP half). Prompt in natural language (Qwen3-VL encoder), not tag soup.SETTINGS (used by every showcase image) — Krea 2 Turbo fp8, euler / beta (simple also fine), 8 steps, CFG 1.0, empty negative through ConditioningZeroOut, resolution 1K-2K in multiples of 64.Every gallery image has the full ComfyUI workflow embedded — drag the PNG into ComfyUI to reproduce it exactly. Stock/native nodes only (ComfyUI 0.26+), no custom nodes needed.TRAINING — ai-toolkit (ostris) on Krea 2 RAW, run on Turbo for inference. Rank 16 / alpha 16, 2000 steps, 75-image hand-curated fantasy-illustration dataset with vision-model natural-language captions. Characters deliberately excluded from the dataset so the style doesn't drag faces or builds into your generations.PAIRS WELL WITH — brushcelstyle, my painterly semi-3D LoRA: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2749544/brushcelstyle-painterly-semi-3d-style-krea-2">https://civarchive.com/models/2749544/brushcelstyle-painterly-semi-3d-style-krea-2</a> — chain both LoraLoaderModelOnly nodes and use both triggers: dndstyle 1.0 + brushcelstyle 0.7 for illustration-first, 0.7 / 1.0 for painterly-first, 0.8 / 0.8 balanced. The combined shots in the gallery are labeled with their strengths.License: free to use, merges allowed, commercial image generation allowed. Credit required; please don't resell the model itself.

Files with SHA256: cd147120ecd23f8e63d1674ed7461de32886b31255a6792913b6e72ef1def222