I have opened up a request form
Sesame Akane from UberQuest. Works best around 0.8-1.0 weight. Training done on NAI.
Note as this was trained on NAI it will not work with Fluffyrock based models.
Trigger: "SesameAkane, colored skin, furry, snout, cat ears, short hair" always needed, for accurate physique add "toned" and "abs" where they would be visible.
Suggestions/notes:
Disclaimer: Works well on most Models I tried but I cannot say if it will work on all. Skin fails quickly as you lower weight so don't complain if it looks off on 0.8.
"Cat tail" in positive prompt if you wanna try fighting SD otherwise put in negative to avoid it.
Outfit 1 (Early chapters): "outfit1, bob cut, red scarf, open vest, brown vest, green shirt, crop top, sleeveless, midriff, abs, green shorts, green vambraces"
Note the intended hair style is "bob cut" for this outfit.
The "green vambraces" is a bit random due to data.
Outfit 2 (Mid chapters): "outfit2, bob cut, red scarf, green armor, green breastplate, single shoulder pad, teal skirt, yoga shorts, green vambraces"
Note the intended hair style is "bob cut" for this outfit.
The "green vambraces" is a bit random due to data.
The "teal skirt" is a bit random due to data.
Outfit 3 (Later chapters): "outfit3, spiked hair, red scarf, grey shirt, sleeveless shirt, bandaged arm, chest harness, belt, high-waist skirt, blue leggings, single shoulder pad, shorts under skirt"
Note the intended hair style is "spiked hair" for this outfit.
This outfit is very unstable due to poor data even from scrapping the comic.
Flexible clothing wise even at 1.0 weight as long as prompt tags do not overlap with outfit tags.
Check samples above for ideas.
Loved this comic back when I found it ages ago and who doesn't like a fit cat with some nice abs. I spent a long time trying to fix issues with my trainer before I could make more loras. I also ended up having medical issues hold me back for the past 2 weeks but now I am slowly getting back in the swing of things. I hope you all enjoy.
Feedback and reviews are always appreciated.
Nerdy training numbers:
Trained on D8Dreambooth trainer
Optimizer: AdamW
Training resolution: 512
Unet LR: 0.0001
Tnec LR: 0.00001
Unet weight decay: 0.016
Tenc weight decay: 0.020
40 Epochs - 6880 Steps
Trained on 171 images no reg.