<h3 id="heading-421">Isaiah, the Sheperd of Chaos from King's Raid</h3>trained on Pony Diffusion, strength varies from 0.7-1.0, recommended strength is 0.9, lower accuracy if you're trying to make something specific in term of outfit/view/etctraining done locally with Prodigy, 8 epochs, 17 images, 5 steps for a total of 704.trigger word is isakr, gradient hair is required.additionnal tags:<ul><li>white strapless dress, wide sleeves, veil, detached collar, white half gloves, white thigh boots, high heel boots, pelvic curtain, cleavage - her "default" outfit</li><li>gradient eyes, purple eyes, yellow eyes, bright pupils - more eye accuracy (doesn't get them right all the time tho)</li><li>large breasts/huge breasts - no comment.</li></ul>negative tags:<ul><li>symbol-shaped pupils, heart-shaped pupils, grey hair - help with getting her eyes right and hair more white than grey (it really is white but SD seems to think it's grey)</li></ul>few things:/!\ Even though I wanted to train it for SD1.5 as well, it couldn't get it right over a few tries so it's only useable with models that are based off PonyDiffusion XL V6 /!\ (I'll try to port it with Animagine when I have time for people who are used to the non-ponyXL workflow) I was honestly surprised with how well the LoRA went seeing that the dataset was one of the worst I've ever worked with. special thanks once again to @<a target="_blank" rel="ugc" href="https://civarchive.com/user/novowels">novowels</a> who patiently answered my questions about PonyXL training on discord dataset: 0 fanarts, 2 images from the official concept art(including crops), 6 images from in-game illustrations(including crops), 9 images from in-game screenshots(including crops)

Files with SHA256: f4023b0bd49256f4b4f88a08602e967b217bf46681f2e51ba916b716982fd5e0