<h1 id="satoru-wada-art-style-lora-for-anima-3.8b">Satoru Wada Art Style LoRA for Anima 3.8B</h1>A style LoRA capturing the style @satoru_wada<hr />Note: Trained on Delta Mix Anima (3.8B). I have not tested it on other Anima 3.8B models, so results on different checkpoints or base versions are not guaranteed.<hr /><h3 id="usage-settings">Usage Settings</h3>Base Model: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2896329/anima-delta-mix-38b-qwen-06b?modelVersionId=3405806">Anima delta mix 3.8B</a>Trigger Word: <code>@satoru_wada</code>Recommended LoRA Strength: <code>0.85 - 1.05</code> (1.0 is the sweet spot)Sampler: <code>euler_ancestral</code>Scheduler: <code>simple</code>Dual-Pass CFG: Pass 1: <code>4.5</code> (11 steps) | Pass 2: <code>3.6</code> (15 steps, denoise <code>0.5</code>)<hr /><h3 id="recommended-prompt-structure">Recommended Prompt Structure</h3><code>@satoru_wada, safe, colossal dragon ...</code><code>@satoru_wada, safe, 1boy, knight, heavy armor, holding sword, facing colossal monster, giant serpent dragon ...</code><hr /><h3 id="training-details">Training Details</h3>Dataset: 50 hand-curated, high-resolution illustrations tagged with Danbooru tokens.Architecture: Anima 3.8B LoRA (Rank 32, Alpha 128).Epochs: 10 epochs (500 steps total), trained with AdamW.<hr />Disclaimer: This model is intended for personal and non-commercial artistic exploration to study the visual style of @satoru_wada.

Files with SHA256: b91ee8d63c3aa3f3869654952fd4442da3eea6aba72fa45cae6ef412dc8a20f6