SDXL - Covered/Veiled In Stuff版本v1.0 (ID: 518829) 综合资源合集综合资源合集

Expressionistic artistic photography of people covered and wrapped in all sorts of translucent, semi-transparent, opaque material and substances... or whatever it's called. Expressive poses included. Some have no people at all in them and me prompting for a particular material doesn't mean it was trained on it, just wanted to see what happens. ( the list below for some triggers ).

While searching for images, i get a lot of weird recommendations from the sites i'm searching on. Of course, me liking weird things i went straight at it. Saw those pictures as an opportunity to at some randomness to my current, never ending , projects. At the moment pretty much in Monster-Creature-Mech mode.

I see all LoRA's, mine or everyone else's, as an enhancement to certain things, even if it misses the theme completely. Some add quality to bad looking images, some add colors to dull pictures, some add details etc. even without excessive prompting, which seems quite popular at the moment ( or necessary 'cause there isn't much room for testing using on-site generator )

----

Based on the source material, has tendencies to make everything nude. There is no picture that really shows explicit nudity in the dataset, but a lot of bare exposed skin, female/male shaped things and body outlines. Of course most models just eat that shit up alive and produce all kinds of raunchy stuff, so be aware if that is a problem for you.
Unfortunately, it also produces a lot of blank and simple backgrounds. Even prompting for it is tricky and will end up negating the main effect in some cases.
I generated quite a few pictures across several models. Tried to keep as simple as possible for showcase reason, but i will post the other pictures later, like always.
After my little "Pimp Bee Buzz" set, i got quite fond of a plugin for A1111 that's called "CADS", which i've installed a long time ago but never used. Have to say thou, for me personally it's a gold-mine, 'cause it adds so much unpredictable stuff. Used it in a few pictures. It's not necessary, just saying.
Last pictures are made using Pony. I know that LoRA's that aren't trained on Pony have little to no effect, but at higher strength they have. Used another LoRA that i've trained on Pony before it got so popular along with this LoRA just to see what happens. There aren't many models out there that really differ from other models, even the ones that claim to be trained ( to what extend, unkown ), but easy to figure out using the same seed, settings and prompt. Of course, that depends if it's a general model or some that is aiming for a particular style or subject. Of the like 50 models i have right now, only a hand full of SDXL models, Pony and Anime models actually do something different to a higher extend.... just an unrelated side note.
Only like 32 images in the set. Main triggers are covered in and wrapped along with colors, translucent/semi-transparent. The main material was fabric ... tried to keep the captions as simple as possible, 'cause the more you add the more you need to use them to make it work. Some pictures have no captions at all because i had no idea wtf is going on, so i left them blank.
As expected, a few of the showcase pictures got tagged "R" contend, so with filter on there could be some missing.

A lot more that i could say, but i let pictures talk. It does what it was trained on and it might be a bit situational, but in the end it's just a little "enhancer" and something to fool around with for me.