An attempt at making a "style"-type LoRA. Definitely not a 1-to-1 copy of the artist's style, but the result is decent enough, I think.Trigger word: magmadoughTested at weight 0.8, but other values may work as well.<h3 id="about-the-versions">About the versions</h3><ul><li>V1 - initial training</li><li>V2a - much less artifacts, but features a less pronounced style. Best used with any other LoRAs.</li><li>V2b - may artifact, but is closer to V1 visually. Behaves a bit like an overtrained model, so you might not get the best result on your first try.</li></ul>

MagmaDough style - Illustrious

Trained on AI-Toolkit instead of Kohya scripts (but still via Civitai trainer). Now it doesn't look overtrained.

Files with SHA256: fa0c930c1c9fe2d40b99817f3b427e4eb7922149e423fd78ae95af82966c8218