<h2 id="lora-based-on-agm-:-https:x.comagm86997980">LoRA based on AGM : <a target="_blank" rel="ugc" href="https://x.com/hachifukuDZ">https://x.com/AGM86997980</a></h2><h2 id="trained-using-wai-il-v15">Trained using WAI IL V15</h2><h2 id="trigger-words:"><span style="color:rgb(250, 82, 82)">Trigger words:</span> </h2><pre><code>agm stylerx</code></pre><h2 id="recommended-weight:-0.7-0.9"><strong>Recommended weight: 0.7-0.9</strong></h2><p>Note:</p><p>This is trained with mixed sfw and nsfw dataset, excels in nsfw AGM style</p><p>This style work best with plump body(Style bias)</p><p>Lower weight tend to have better background, you might lose the style the lower you go.</p>

[Style | illustrious-XL] AGM

<p>Re-trained with Training Revision 2</p><p>RX Training Revision 2(TRev2) training introduced better image ratio handling, fixes object out of frame issue and better background handling</p><ul><li><p>Color contrast is improved, less grayish compared to previous versions, but sometimes it <strong>may over contrast</strong></p></li><li><p>More aligned with artist style, compare to RXS2</p></li><li><p>It should be less prone to lora style overtaking</p></li></ul><p><s>Some how this lora becomes the flagship of all my loras, interesting...</s></p>

Files with SHA256: db78f82acd4a4a5cdb67c5a053e0ff42f13a768dd55f20ef4ce791fe54561e7b