try any of these - or put in Neg ( when altering appearance/outfit)
may need to add 3d, or realistic, in neg for certain 2d/anime styles - lower weight for less 3d influence - 0.8 between 0.6 - may have more 3d influence
Pony XL - May be more accurate but have more 3d influence
ILL - same TW
face/body
alterxl, white eyes, black sclera, green hair, or aqua hair, short hair, or low ponytail,
dark lips, or lipstick,
black face paint, or black facial mark, - put in NEG to reduce
may have scars, or body scars,
face paint, may add tattoos, on body
Attire
white bodysuit, or bodysuit, black glove, single prosthetic arm,
clothing cutout, pants,
pink mechanical tail, black mechanical horns, single mechanical arm, (left arm)
if removing tail, suggest mechanical tail, floating objects, in NEG - Also may spawn two tails. twintails, or twin tails, IN NEG may help
may try to hold tail suggest holding, in NEG if so
attire is specific so some prompts are missing or inaccurate - was also trained without gear - so neg use may be useful
science fiction, may enhance
Description
FAQ
Comments (7)
hi, in the meantime I would like to try to create a Lora Pony XL, how do you do it? I use: Lora Trainer by Hollowstrawberry for the Lora sd1.5
when you go to the colab of 1.5 by Hollowstrawberry - on the top left there is a link to a ponyxl/sdxl lora trainer page - though may need to pay for colab plus 10$ if planning anything more training wise
@True_Might hi, yes at the top left it says: XL Lora Trainer. But it doesn't fit: dataset maker XL, why? To create the sd1.5 I do the first step on dataset maker, then I move on to the second step i.e. Lora trainer. But what should I do to create a Pony Lora XL?
@ApexThunder you go here https://colab.research.google.com/github/hollowstrawberry/kohya-colab/blob/main/Lora_Trainer_XL.ipynb and do the same method you do to make 1.5s but just with sightly different settings. You can even use the same 1.5 dataset of any character or style to make a pony lora - XL loras just take more resources to train.
you should be asking the folks who made it - I just make loras I dont do all the coding and whatnot
@ApexThunder just make the dataset in dataset maker colab - and instead of putting it in the 1.5 lora trainer - put it in the XL lora trainer
@True_Might hi, so I have to start from: Dataset Maker and then move on to XL Lora Trainer?
@ApexThunder yes - if you think theres some kind of XL dataset maker there isnt one