A test chamber, a simple concept really. People, animals, monsters, you can put all sorts of critters in it and observe them from the safety of the protective glass.Trained with 51 images on the PonyDiffusionV6 checkpoint. The sample images were generated with AutismMix and Everclear PNY.Would love to see what you guys can do. Feel free to post any pictures here.Recommended settings:<ul><li>Sampler:DPM++ 2M Karras - 25-35 steps</li><li>CFG: 7</li><li>Resolution: 1024*1024, 1024*1280 (See notes below!)</li><li>Lora weight: 0.8 - 1</li><li>HighRes: 4x_NMKD-Siax_200k, upscale by 1.5x, 20 steps, 0.35 denoising</li></ul>Basic prompt:testchamber, glass wall with text "TEST SUBJECT",Useful tags:<ul><li>Clean theme: clean, bathroom (put the horror theme's tags in the negative prompt)</li><li>Horror theme: horror, dirty, rusty, broken neon light</li><li>Anime/Cartoon style: source_anime / source_cartoon, anime screencap (put "3D" in the negative prompt)</li><li>Closeup shots: close-up (more detailed characters)</li><li>Full body shots: full body (less detailed characters)</li></ul>Additional notes:<ul><li>The training images were all 3d renders so if you want something stylized you have to put "3D" in the negative prompt.</li><li>Works best with square aspect ratio. Will work with portrait and landscape but the text will get messed up if the width of the image exceeds 1024px. The two recommended resolutions above will keep the text intact most of the time.</li><li>The text is unstable, that's just how SD is. If it bothers you, avoid mentioning "glass wall" and "text" related tags in your prompt.</li><li>Don't forget about the Pony's usual quality tags.</li><li>My samples have the prompts embedded.</li></ul>

Test Chamber

testchamberV2.safetensors

V1.2

A test chamber, a simple concept really. People, animals, monsters, you can put all sorts of critters in it and observe them from the safety of the protective glass.Trained with 51 images on the PonyDiffusionV6 checkpoint. The sample images were generated with AutismMix and Everclear PNY.Would love to see what you guys can do. Feel free to post any pictures here.Recommended settings:<ul><li>Sampler:DPM++ 2M Karras - 25-35 steps</li><li>CFG: 7</li><li>Resolution: 1024*1024, 1024*1280 (See notes below!)</li><li>Lora weight: 0.8 - 1</li><li>HighRes: 4x_NMKD-Siax_200k, upscale by 1.5x, 20 steps, 0.35 denoising</li></ul>Basic prompt:&lt;lora:testchamberV2:0.8&gt; testchamber, glass wall with text "TEST SUBJECT",Useful tags:<ul><li>Clean theme: clean, bathroom (put the horror theme's tags in the negative prompt)</li><li>Horror theme: horror, dirty, rusty, broken neon light</li><li>Anime/Cartoon style: source_anime / source_cartoon, anime screencap (put "3D" in the negative prompt)</li><li>Closeup shots: close-up (more detailed characters)</li><li>Full body shots: full body (less detailed characters)</li></ul>Additional notes:<ul><li>The training images were all 3d renders so if you want something stylized you have to put "3D" in the negative prompt.</li><li>Works best with square aspect ratio. Will work with portrait and landscape but the text will get messed up if the width of the image exceeds 1024px. The two recommended resolutions above will keep the text intact most of the time.</li><li>The text is unstable, that's just how SD is. If it bothers you, avoid mentioning "glass wall" and "text" related tags in your prompt.</li><li>Don't forget about the Pony's usual quality tags.</li><li>My samples have the prompts embedded.</li></ul>

Files with SHA256: 5f1f512848319bf5d0e96d44621db78275568787bd7c911ea2e81f7f3b1c89d7