This model generates a catgirl named Hart. She's a lover, not a fighter.She was developed from a single image, specifically <a target="_blank" rel="ugc" href="https://civarchive.com/images/21965873">this one</a> (NSFW, rated R by Civitai). I used iterative signal amplification to generate more data of her and create a better model.The trigger word is <code>hart catgirl</code> and that's all you really need for it to work. <code>pink eyes</code> and optionally <code>heart pupils</code> might also be good if you're dissatisfied with her eyes. Don't expect accurate heart pupils though, they're more like pointed at the bottom with a prism/fractal gradient pattern.Trigger word: <code>hart catgirl</code>Optional helper captions: <code>hearts, heart hairband, heart earrings, heart pupils, pink eyes, pink hair</code>I'm marking this model as mature because the model is super horny and really just wants to make NSFW content.One of the things I like about this model is that it also wants to incorporate hearts into every possible element related to her. She usually has a cloud of hearts around her and her clothes and held items often have heart elements incorporated into them. Considering the original image was just repeating the word "heart" over and over, this seems fitting.This model was regularized on my <a target="_blank" rel="ugc" href="https://civarchive.com/models/524110/oatd-style">Oatd</a> and <a target="_blank" rel="ugc" href="https://civarchive.com/models/422449/galena-bikini-pinups-for-pony">Galena</a> datasets, so it might have some bias from those datasets. This time I marked the data with <code>oatd</code> and <code>galena</code> respectively and the model responds to those words, which is especially noticeable with <code>galena</code> which produces a much more semi-realistic style. The base style is the same as the original image, which came from <a target="_blank" rel="ugc" href="https://civarchive.com/models/498713/perfectrsbmix-pony-real-anime-sfw-nsfw?modelVersionId=662611">this model</a>. This model itself has been trained on vanilla Pony and will give the most faithful results using that base model.Works best at 1024x1024, 1024x1536, and 1536x1536 resolutions.The people of Civitai seem to love my terribly unbalanced draft work so I've decided to release this one before it's fully complete. The first version only has 3 images in the dataset, but it actually works surprisingly well and the only real downside is that it's wildly unbalanced, and honestly it's pretty fun to play around with.

Hart Catgirl (OC)

Retrained at Rank 128 with sv_fro 0.9 resize (current best practice) and added 2 new reference images. This one is nearly v1.0 but I want to add a few more training images.

Files with SHA256: f28818a93f3dd2157acf327b6cb4f6aabeaf45346807615d4dfcc3ba04b16dcc