Kitty Softpaws (Puss in Boots: The Last Wish)

v3.0

Trained on the same dataset I used for Pony XL LoRA, but this time I added more captions.

30 cropped screenshots from the original movie with all other characters removed.

Have fun and share your art below! ❤️

v2.0

My personal achievement and best LoRA ever!

Use 'KittySoftpaws' as a main trigger word.

Use 'green bandage on waist' and 'boots' for movie look.

DPM++ 2M Karras

AutismMix SDXL for better anatomy and poses

v1.5

I trained a new model on 800 screenshots without cleaning them from other characters. Then I merged v1 model with it with 0.9-0.4 weights.

So now it has better anatomy understanding and full body details.

BUT I recommend using AutismMix Pony model for better anatomy and better fingers.

Also it seems that this version has better emotions and require lower LoRA weight: 0.6-0.65 is enough.

BASE PROMPT:

score_9, score_8_up, score_7_up, score_6_up, rating_safe, source_cartoon, source_furry, (ksp:1.2), black anthro female cat, solo, two tone fur, big eyes, (<emotion>:1.2), <other tags>

'full body, bandage belt, leather cowboy boots'

BASE NEGATIVE:

cat eyes, big ears, rating_explicit, nude, breasts, nipples, source_pony, score_4, score_3

Generation settings are same from v1. Weight 0.65 is recomended.

I hope you enjoy using my model and you will support me with buzz!

v1

I didn't find a high-quality LoRA for Kitty Softpaws from Puss in Boots: The Last Wish movie so I trained my own. I used 119 unblurry screenshots from the movie, each 1024x1024. I've removed any details from other characters that accidentally hit the frame. Mostly close-ups of the face and emotions were taken, but I also included side and full-length views.

I added simple tags like 'full body', 'bandage belt' and 'boots' for full body, 'upper body' for upper body and 'close up' for face shots. The main trigger-tag is 'ksp' (kitty softpaws), but I can say that 'female cat' is necessary.

Try to add '(ksp:1.2), solo, anthro female cat' as a default, it seems to have a closer look to the original character.

For full body shots with original clothes use 'full body, (bare thighs:1.2), bandage belt, boots' in prompt and '(pants, skirt:1.3)' in negative.

Both Euler a Karras and DPM++ 2M Karras looks good, maybe DPM is slightly better in details. 30 steps is enough to create a detailed fur and eyes, but with 40 steps you can get a more detailed result.

By default, the emotion is serious, so you may need to include it in a negative list if you want a highly positive face. There were also few images of her tail, so you can include 'thin cat tail with white tip'

Use Clip skip 2 and 0.8-0.85 LoRA weight.