<h2 id="update-ahyjpjxn9">#Update</h2>Makes you think, the model I made for SD15 was almost 2 years ago... anyone left here from that time?Some info about Illustrious version. It's a bit of a pain to train, since most images are less than 500px (both height and width usually).I tried training two outfits:<ol><li>The camisole and pants, where the trigger is -- j3n_c@mi</li><li>The cheerleader outfit, where the trigger is -- j3n_chr</li></ol>Obviously, you might need some tags describing the outfits themselves, like purple camisole, white pants for the first one and blue shirt, white skirt for the second (also, I wouldn't expect the P letter print on the shirt to work consistently)Weight is fine up to 1.0, but it can mess up the fingers sometimes, if that happens, lower it slightly.I might fix it more in the future, but not right now, since I've already made more versions than I'd like to admit, looking for a good result.<hr />Anybody remember this cartoon?, good times.Here's Jen from the show 'What's with Andy?'. I wanted to do something from a more dated show, and it really is. The dataset was rather bad, so I'm surprised this actually worked out. For the prompts:The trigger word is jen Useful tags are: parody, purple shirt and spaghetti strip for the camisole she's wearing, black eyes, red hair, and white pants.If she gets freckles, tag for freckles in negative prompt. * I didn't shuffle the tags, so moving them around in the prompt can have a somewhat big effect, you can experiment with that. I do recommend putting the character lora and trigger right at the beginning of the promptI tested the weights in the range of 0.8 to 1.0, it depends on what you want to do. If you use other loras with it I would recommend dropping a bit.

Files with SHA256: e0383a7c0926ebb826950067e0be0e8eeadf68ba6b36d177fa9fd37ba94f2220