A style LoRA for Qwen-Image trained on 244 curated illustration and anime reference images across diverse visual styles. Produces images spanning cute anime character design, European bande dessinee, indie risograph prints, storybook watercolor, retro shoujo manga, and graphic novel illustration.<h2 id="style-influences">Style Influences</h2><ul><li>Japanese anime and manga illustration -- clean cel shading, expressive character design, shoujo and slice-of-life aesthetics</li><li>European graphic novel and bande dessinee -- ink crosshatching, atmospheric landscapes, ligne claire</li><li>Indie illustration and zine culture -- risograph print textures, limited palettes, grainy halftone</li><li>Children's book and storybook illustration -- watercolor washes, warm palettes, charming character proportions</li><li>Retro anime aesthetics -- 80s/90s anime film stills, VHS grain, bold color blocking</li><li>Concept art and character design -- flat color fills, turnaround sheets, fashion illustration</li></ul><h2 id="usage">Usage</h2>No trigger word -- this is a style LoRA. Describe what you want and add style cues like "ink and watercolor", "clean cel shading", "risograph print" to steer the output.<h3 id="recommended-inference-settings">Recommended Inference Settings</h3><pre><code>Sampler: euler
Scheduler: simple
CFG: 3.5
Steps: 45 (30-60 works well)
LoRA strength: 0.8-1.0</code></pre><h2 id="training-details">Training Details</h2><ul><li>Base model: Qwen-Image (FP8 quantized, text encoder FP8)</li><li>Training steps: 59,000</li><li>Rank/Alpha: 42/42</li><li>Learning rate: 5e-5</li><li>Optimizer: AdamW 8-bit</li><li>Caption dropout: 0.35</li><li>EMA: enabled (decay 0.99)</li><li>Dataset: 244 curated images across 4 subsets, trained sequentially then consolidated on the full dataset</li><li>Trainer: <a target="_blank" rel="ugc" href="https://github.com/ostris/ai-toolkit">ai-toolkit</a> by Ostris</li></ul>

Files with SHA256: a196b3f13321edfe116f8b6d0c191f4012dabd4bf83060109aa52ffff3aeaa1c