Would appreciate comments, posts, reviews, and requests
if you'd like to pay for a commission or just send a dollar or two -> PAYPAL or Kofi
a Lora for an ass support pose.
0.6 - 1.0
recommended to highres. fix with my other Loras upscaling also seems to help. (so if a generation is slightly off instead of altering prompt or changing seed try upscaling)
recommended width = 512 height = 768
recommended tags: ass_support
optional tags: from_behind, from_below
V7: This version I got rid off all variants and it is strictly 2 handed ass_support, also added a few dozen images.
WITH OBJECT:
it is a bit hit or miss, the items i used were the ones in the dataset, not sure how well other items will work.
I would appreciate any feedback, images you make with my loras, and requests in the comments.
haven't tested in conjunction with this Lora Grabbing own Ass | Stable Diffusion LORA | Civitai
think V5 is undertrained and the characters from the dataset may be baked in
also V5 requires LoCon and Lycoris extension to work
Description
well, guess it's back
LoCon and Lycoris extension required to work
FAQ
Comments (7)
@sorza
Actually, trigger words are a great thing. because of two reasons:
1. if you have a trigger word you can put that in the negative prompt while putting the lora itself in the negative prompt won't have any effect.
2. it allows for using advanced methods such as invoking the Lora after a certain number of steps like so: "apple [triggerword:20]" which tries to apply the style of the Lora to the apple only after step 19.
and sometimes having a basic trigger word can mess up the training as well, or if there is no trigger word it is harder to use multiple Loras, etc. it is just easier with them unless your doing a style
for 1 you can use lora with negative weight in positive prompt. For example,<lora:4chanBBWStar_4chanBBWStarV1:-0.2> makes characters skinny
On the other hand, if that LORA has been taken down (like we see often on this site) you can't look up what the trigger word is, and thus can't use the LORA properly at all.
It'd be nice if the filename had the trigger word in it, or there was some other way to figure out what it is, cuz some of these LORAs have the most un-guessable trigger words you've ever seen.
@namelessoptimist I see your point, but none of my models have been taken down, I try and follow the guidelines as much as possible. will probably take your advice though, seems like a good idea to name the file the trigger word
@namelessoptimist just make notes in .txt files per lora ez
@namelessoptimist Most LoRAs, you can view the LoRA info to see how many of the images had what keywords, and deduce from that "there's 236 training images and the word 'buttlift' is tagged 236 times, so its the trigger word"