D2PonyXL v003d

4 trainings with 99384 total Steps

Dataset size: 1782img for training a+b, 929img ("best of") for c+d

ca 70h training on Prodigy O_o

Intended for mixing with your promts for curvy body-shape analog photo-look backgrounds. Use at <lora:FILENAME:0.66> or <lora:FILENAME:0.3:0.8>for anime-look, everything above will give you realistic results (Basically the same deal as it was with SD1.5 version i posted before.)

Use low denoise 0.1-0.2

Recomened rez: 672x1024 or 1024x672

trained look concepts:

dim lighting

warm colors

bloom

analog photograph

foreshortening

there are several trained haircuts, shiny skin, etc (see tag freq in metadata).

"large breasts, large areolae, precise tan" are mandatory ;)

recommenced checkpoint:
Pony Diffusion for Anime