Presenting my third Overwatch LORA. With this one I focused on trying to package 3 distinct outfits - original, Cote D'azur and Talon skins. Bunch of instructions down below.
Checked strength range is 0.6-0.8.
Main triggers for prompting skins are:
Original - simple prompt widowmaker should sufice, with perhaps an added visor. widowmaker IS NOT required if you specifically prompting for two other skins, it's addition may have varying results.
Cote D'azur - cote d'azur, blue bikini, blue skirt, choker, hoop earrings, purple skin, blue skin, front-tie top, swimsuit, sarong, eyewear on head, sandals, red lipstick - feel free to add\drop things from list as you need.
Talon - widowmaker_talon, beige skin, talon armor, black pants, jacket, midriff, navel, talon alternate costume - a bit weaker then other two in terms of replication and detail, but still should give a distinct look.
Original skin rarely might just slightly bleed into other clothes - easily dealt with by adding bodysuit and visor into negative.
Triggers for paticular things are:
Outfit - bodysuit, pauldrons, gauntlets, gloves, armored boots, high heels
Tattoo - back tattoo, emblem, animal print, sadly you most likely won't be able to get exact tattoo even with inpainting, Widowmaker tattoo is too tricky for a three-skin dataset
Accesories - visor, head-mounted display, talon eyewear on head (often just puts sunglasses on, so add sunglasses and glasses into negative)
Gun - holding weapon, sniper rifle, assault rifle, talon rifle, scope, aiming (specifying in which direction might help), firing
Review and comment with your experiences, I would appreciate it.
Making and testing LORA's takes quite a bit of my time, so I would appreciate if you bought me a coffee.
Trained on Anything V4.5 with anime.vae.
Samples are 512x768\768x512, not cherry picked, without HighRes.fix, using OrangeCocoa 50/50 Mix , with OrangeMix VAE, and using EasyNegative embedding, CLIP Skip 2. Inpaint examples are txt2img with 1 single inpaint done on eyes (I will comment on exceptions).
Dataset of 60 images with deepbooru tags
Go ahead and upload yours!
Your query returned no results β please try removing some filters or trying a different term.