<h2 id="heading-32">Photo styles of Animagine XL V3.1</h2><p>The model is trained based on Animagine XL V3.1 as the foundational model, with materials derived from character images generated by Animagine XL V3.1 using real human models combined with controlnet for material generation. Animagine XL V3.1's prompt can be utilized.</p><p></p><p><strong>Prompt Wildcards</strong>： <a target="_blank" rel="ugc" href="https://civarchive.com/models/392957?modelVersionId=438325">https://civarchive.com/models/392957?modelVersionId=438325</a></p><p></p><h3 id="heading-2232">Tag Ordering</h3><p>For optimal results, it's recommended to follow the structured prompt template because we train the model like this:</p><pre><code>realistic, 1girl/1boy, character name, from what series, everything else in any order.
</code></pre><p>The anime characters are too exaggerated, If you encounter the problem of eye collapse, so you need to reduce the cfg, you can try the following settings:</p><p>CFG scale: 2-5.5</p><p></p>

Files with SHA256: 9e924bbe2434e985107cf0ef84cc5b58cfe96f46f14a3694922295bc89191637