Illustrious version of this LoRA here: <a target="_blank" rel="ugc" href="https://civarchive.com/models/1654243">https://civarchive.com/models/1654243</a>Huge thanks to <a target="_blank" rel="ugc" href="https://civarchive.com/user/nl196">nl196</a> for his commissioning of this LoRA, as well as his help in gathering the images necessary for the training! This sentai costume is really good!Notes:<ol><li>Due to the training data set consisting entirely of photos, the faces were cropped out to prevent the AI training on the faces. As a result, AI might struggle to produce images without helmets being worn. If generating such images, would advise adding "helmet, wearing helmet" with sufficient weightage to negative prompt, and then hair prompts with sufficient weightage to positive prompt</li><li>Likewise due to point number 1 above, you can't use "hand out of frame" because AI seems to confuse it with "head out of frame" and crop off the face lol</li><li>For some reason, there are two different available designs with the chest symbol, one being wing-shaped and one being M-shaped. Images of both were added into the training data set and captioned with "wing-shaped chest symbol" and "m-shaped chest symbol" accordingly, but doesn't seem like AI figured out what is going on here from my testing LOL</li><li>I captioned "holster" for the handgun container attached to the belt and "holstered" for if the handgun is inside the container, but nope, seems like AI is mixing both terms up and not understanding what "holstered" means</li><li>For some inexplicable reason, DHRY-11 uses a latex costume instead of the others using spandex costume, so they were captioned accordingly. Use whichever you prefer (I myself prefer "spandex" haha)</li><li>Increment the weight of the "spandex" / "latex" prompt if it is not shiny enough for you!</li></ol>Prompts:Main prompt:<pre><code>sentaisaintforce, [specify the color you want here!] bodysuit, silver belt, white gloves</code></pre>For full body images:<pre><code>boots, white footwear</code></pre>Material of costume (there were two different materials available in the training data set):<pre><code>spandex</code></pre><pre><code>latex</code></pre>Symbol that appears on the chest (there were two different materials available in the training data set), though from my testing the AI doesn't seem to have figure out what this prompt means and just generates whichever based on its fancy LOL:<pre><code>wing-shaped chest symbol</code></pre><pre><code>m-shaped chest symbol</code></pre>Weapon prompts (second is meant to have a handgun in the holster, but AI would just generate an empty holster and the handgun in hand so *shrugs*):<pre><code>handgun, holding gun</code></pre><pre><code>holstered, handgun, holster</code></pre><pre><code>holster</code></pre>Helmet prompts (first for wearing it, second for not wearing but holding it, and third for it being absent in the image at all. If going for the third and helmet is spawning, can add "helmet, wearing helmet" to negative prompt):<pre><code>helmet, wearing helmet, [specify the color you want here!] helmet</code></pre><pre><code>helmet, helmet removed, no headwear, unworn helmet, holding helmet, [specify the color you want here!] helmet</code></pre><pre><code>helmet removed, no headwear</code></pre>

Files with SHA256: 825eb9ecc6104baa3f85e8423687b2e70fe9b5489cbe723b26ae3127153b2031