ABOUT V1.5 UPDATE:
I am trying to make a final "perfect" version of this LoRA and it's going slowly, so here is the "V1.5" version I've been using while working on it that I used to make my last several images of her so everyone else can use it, too. I do think it's vastly improved over the first one (my first LoRA ever), even if not perfect yet.
Main Trigger Word: FreijaOCv2
Helpful Words (adjust as needed): anthro cat girl, two tone eyes, colorful purple hair, pink nose, cat ears, detailed body fur, white fur, long glittery hair, cat tail, masterpiece
Negative prompt: outline, lip, bad quality, worst quality
Because I wanted an emphasis on fluffy body fur (which is very detailed and pretty in this version) the training data does contain a mix of SFW and NSFW in it and can do both. If you do not specify any kind of clothing, she might not have any, so be aware and prompt with care.
Original Version:
This is a LoRA to help creating Freija, my OC/Avatar of myself. Trained on Illustrious. Results vary by base model, high realism obviously isn't a good choice, but it does work nicely with some anime models not anthro by design. 36-steps are highly recommended for good results. To prevent randomly rearranging the tail a low denoising strength is recommended when upscaling- around 0.3 through 0.45 worked well for me.
On semi-real models that add human lips and look weird you can use "lips" or "pink lips", etc. in the negative prompt.
I used only images that I generated myself, nothing was stolen from the net or other artists. The data was varied heavily and seems to produce reliable results across different poses/outfits/expressions.
Please consider leaving a review with an image post! I'd be grateful to see what you create. Feedback, including constructive criticism, is sincerely appreciated!!
Finally, this is my first LoRA and it is dedicated to the amazing @Kayot, who makes gorgeous anthro style helper loras and gave me encouragement when I first decided to create a LoRA. This wouldn't exist without @Kayot's kindness and assistance! ^u^
Description
Improvements over V1:
-Vastly improved face consistency- she actually looks like her in almost every scene.
-Body fur is fluffier and more detailed. She is a fluffy cat, not a shorthair, after all.
-Two-tone eyes have better detail.
-Tail doesn't clip or vanish between legs or behind jackets much, and is less bad about clipping through things.
Regressions from V1:
-Eyes lack color consistency, a bit random.
-Tail sometimes has a two-color tip and not a single light purple section.
-Sometimes she has an outline, like an anime character, adding "outline, illustration" to the negative prompts helps keep it from happening.
FAQ
Comments (2)
This is such a beautiful and adorable character. 💜
I really like how her horn fits together with her feline features - it gives her a very unique and charming appearance. Her purple hair and colorful eyes also make her especially memorable.
I also made a comparison between the two versions of this wonderful character, testing how they differ in appearance and how they handle different styles - https://civitai.red/articles/33570/freijafoxy-oc-review-two-loras-one-character
Thank you again for the detailed and lovely images and review! The v1.5 is a step forward, hopefully a proper v2 can be finished soon- I've got a lot going on, but as my LoRA of me as a creator, she's the #1 priority when I have time! ฅ^>⩊<^ ฅ