This LoRA captures the 3D art style of hooves-art and is not focused on ponies. It performs well with any character, whether human, pony, or furry.<hr /><h3 id="usage">Usage</h3><ul><li>Trigger words:<ul><li>Illustrious v2.0 / Anima v1.0: <code>hoovesart_style, 3d</code></li><li>Illustrious v1.0 / Pony v1.0: <code>hoovesart-style, 3d</code> </li></ul></li><li>Suggested settings:<ul><li>Illustrious / Pony: CFG <code>4–8</code>, <code>20–30</code> steps.</li><li>Anima: CFG <code>4.5</code>, <code>40</code> steps, Euler a / Simple.</li></ul></li><li>LoRA strength:<ul><li>Illustrious v2.0: <code>0.8–1.0</code>.</li><li>Anima v1.0: start at <code>1.0</code>. Try <code>1.2–1.5</code> for a stronger style, depending on the checkpoint and prompt. </li></ul></li></ul>Character compatibility: This is a style LoRA, not a character LoRA. Character recognition depends on your base checkpoint — some models may not reliably reproduce MLP characters. For a closer likeness, use a checkpoint that knows the character or add a dedicated character LoRA.<h3 id="style-comparison">Style comparison</h3>The examples below compare LoRA off, 0.8, and 1.0. Each row uses the same prompt, seed, and generation settings — only the LoRA strength changes. <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/e2333c30-e2c5-4d64-856a-1af0724aa36b/original=true/e2333c30-e2c5-4d64-856a-1af0724aa36b.jpeg" /> Generation settings: <a target="_blank" rel="ugc" href="https://civarchive.com/models/503815/nova-furry-xl?modelVersionId=2943166">NovaFurryXL_ilV180A</a> · DPM++ 2M SDE / Karras · 30 steps · CFG 5.5. The trigger words <code>hoovesart_style, 3d</code> are included in all examples, including those without the LoRA.<hr /><h3 id="anima">Anima</h3>Tested settings: <a rel="ugc" href="https://civarchive.com/models/2458426/anima?modelVersionId=2945208">Anima Base v1.0</a> · LoRA strength 0.8–1.5 · Euler a / Simple · 40 steps · CFG 4.5. For Anima 2.9B / 3.8B, use a loader with Anima block remapping, such as <a rel="ugc" href="https://github.com/INuBq8/ComfyUI-Anima-LoRA-ControlNet-Patch">this ComfyUI patch</a>. Style strength may vary between checkpoints.<hr />The dataset was collected using a custom parser and processed with <a target="_blank" rel="ugc" href="https://github.com/advimman/lama#pretrained-models">LaMa</a>.

Hooves-art 3D Style | Illustrious · Anima · Pony

Some training tweaks and updated dataset

Files with SHA256: ddc3a4be84b33b686dd6d79ba1174267a7a4209d9efe2ec154aff06e1419d1ec