I made a Matt (Aqua Admin Matt) of Team Aqua LoRA from Pokemon. He's known for his muscular body, his personality, and the way he speak (especially to Archie). I made this beefcake dude wth 130+ images with mostly from Google Images and some various image sharing sites. Mostly trained from fanart. <s>Matt also known as bara material.</s>Trigger Words (choose one)<pre><code>aqua_admin_matt, matt \(pokemon\)</code></pre>Settings I used:<ul><li>Weight: 0.7 - 0.9</li><li>Sampling Method: DPM++ 2M SDE Karras</li><li>Sampling steps: 22 - 25</li><li>Resolution: 768x768 or 960x960 (768x960)</li><li>CFG Scale: 7</li><li>Checkpoints<ul><li>Mature Male Mix</li><li>Airfuck's Brute Mix</li><li>AnyLoRA + Pokemon Style LoRA</li></ul></li></ul>

Aqua Admin Matt | Pokemon | SD 1.5

I updated this model. I noticed that in my dataset there's some incredibly bad image that affects the quality of dataset, hence the decrease number of dataset. Despite decreased image quantity, there also image addition that has much higher quality and all images has bein resized. Here's some breakdown of this model update:<ul><li>Increased image resolution up to 4x from original</li><li>Decreased Dataset from 130+ to 120+ images<ul><li>There's still additional new images that has high quality</li></ul></li><li>Increased LoRA size 2x</li><li>Resolution Train Image Increased to 640</li><li>Simplified namefile from Aqua_Admin_Matt-15 to Matt-15</li></ul>

Files with SHA256: 0a8a3e6ed4814ec2948f3f7673012b7f1249be844eb8213edd6d9a6acf70563c