Insert things into holes and stuff.
Trigger words:
insertion pussy
insertion ass
insertion mouth
insertion breastWorks best in portrait (512x768) and bad in square format (512x512).
Good checkpoints for it:
Good supporting LoRAs:
Sample images are cherry-picked and a few re-generated with the same seed, but should all be reproducable.
Changelog:
v2 - Added "insertion breast" as a keyword. The model is a bit unstable, and too anime.
v3-v6 - Some good stuff, but mostly messy and too anime.
v7 - Pretty clean and powerful version. More photorealistic and high quality.
v8 - May be too mellow, not sure yet. Let's see what you can do!
Description
Base version.
FAQ
Comments (18)
This looks AMAZING!!
Oh my. now we can finally make some decent art of women giving birth, alien tentacle invasion, sword swallowing magic, etc. God bless you my friend.
Happy to help!
I have lots of ideas for future improvements. But for now I wanna see what kind of things you can think of and make work :D
@6tZ in terms of future improvements: do you by any chance happen to know the artist henmaru machino? :D
@eurotaku I do now =)
Shout if you have any other suggestions!
Current training data is just real photos.
@6tZ well, corresponding deformations of the body like stomach bulges would be awesome, but no idea if that goes beyond the scope of a single lora
@eurotaku Didn't even remember the name off the top of my head and I don't read manga or any of that but I still have an English copy of The Holes (not sure if that was the correct translation). It's lived in a subdirectory of the windows pictures library since fairly Windows 2000. Can't forget "There's nothing else I could put inside me..."
Insert things into holes. Fairly flexible.
ic wut u did tharr :D
@eurotaku I know you're using it full-time! Post something =D
@6tZ would be glad if i had the time to, but there are other things to tend to first... ;_;
@eurotaku New,, umm,, inspiration material uploaded for you. From v1. V2 is in progress, shaping up very nicely.
@6tZ woah, oscar-worthy performance. consider renaming your lora to putting everything everywhere - all at once :D
I tried everything (RevAnimated 122) with or without other female Loras, and it doesnt work. Just like Multiple Insertion, items rarelly appear, and when so, they are in different places other than...the holes. I tried 512x512 and 512x768. I tried exchanging prompt places. It does work but like 1 in each 10 renders.
Hmm, it's not foolproof for sure. You get lots of images without the stuff nicely. It depends on the item of course. If you try items more commonly used in these situations, you get a higher success rate (and more proper placement).
It certainly needs further training though. Going to iterate on it over time.
It works best with vertical images, so stick with 512x768.
Items in the current training data include:
Banana, Cucumber, Strawberry, Whine Bottle, Lollipop, Carrots (this one doesn't end up very well), Corn on a cob, Apple, Asparagus, Baseball Bat, Pens. See if you get better results with them to make sure that it at least works as it's "supposed" to for now.
@schsch Also, be sure to use the full trigger word "insertion x", not just "insertion". I have removed any mention of the basic trigger word, it's a bit misleading. It needs to know which hole to put stuff in.
Great! Thank you! Gotta test more. Great job indeed! I just imagine this working on my Ragnarok Loras ^_^
@schsch Not sure what those are, but do share some renders if you get any good results!
Compared with other LORA, this degree of freedom is greater and easier to use. The point is that the item will appear, and it will appear as soon as the prompt word appears. You don’t need to add other LORA, and you will win on this point