PLEASE READ AND USE PROMPT FORMAT FOR PROPER FUNCTIONING
A configurable dora for the stuck in wall/object concept from the front. Has some ability to generate objects outside of training data (ex: "stuck in pumpkin/tree"), so be creative! It won't generate as reliably as the trained object (see list below), but it can be useful with a little fiddling.
Mostly focused on generating frontal poses, but has some ability to generate views from behind as well.
Prompt format:
1boy/1girl/etc., solo, stuck in {object}, {optional body position, like "legs up", "bent over", "kneeling"}, {body part(s)} out, {[optional] {color} grommets} {remaining description}
NOTE: Grommets are the rings that show up around body parts sticking out of the object. This tag is not very reliable and works best with the "wall" object.
V1
Trained Objects (non-exhaustive list, just most common options)
{gold/wood} plaque on {color} wall
{color/descriptor} wall
{color} box
red tentacles
{color} plinth
{color} ball
squat toilet
trough
need to put "squat toilet" in negatives
urinal
not very well trained, will need to put "squat toilet" in negative
may also need to include "cum pool" or "pee pool"
Body parts
head
feet
calves
ass
chest/torso
use one or the other
thighs
hands
To avoid text, put "nameplate" and "name" in negative prompt.
Put body parts in negative prompts if they are showing up even if not tagged.
Not all body parts work well with all objects. Generally, body part and object pairs in sample images work best.
Description
minor improvements
FAQ
Comments (7)
can you do girls stuck in the painting/mirror
I've posted examples of both to the gallery now. It might take a few minutes to show up.
I don't think I had any examples of this kind of scenario in my training data, so it doesn't work as well as other objects, but it's also not too hard to generate.
One of the best concept LoRas I've ever used. You did an amazing job! Any plans on updating it further to add more stuff to the list in which characters can be stuck in?
I'm still collecting images in the background, but haven't seen many other objects I don't already have in my dataset. (not everything in my dataset is listed, since many of them have few examples). For objects not included, some of them still may work to varying degrees, like the "stuck in painting" someone else requested.
@out_of_username_ideas sorry for the late reply, but quick question. If I find any stuck in painting images, can I send them to you through pm? I like this model a lot, and would like to contribute to it.
@YammyDark sure. any novel "stuck" locations would be great. Probably won't do another training run until I have a large number of new images collected though, or I hear about a better training method
@out_of_username_ideas Nice! I'll be on the lookout for them.



















