So, I go on a quick vacation and unbeknownst to me a new model pops up that happens to be named after a psychological concept I find quite interesting (and ironically had mentioned before I went on said very short vacation).
Color me intrigued. Said intrigue aligns with an experiment I was conducting - true to my name, I boiled some good old kompots (or compotes for the head scratchers). The concept is simple -- I pick two (or more) art or artist styles that my ignorant, unquestionably subjective eye finds complementary or fitting, mash them together in a jar, boil the air out of it and present the vacuumed fruity drink for consumption at your convenience (I recommend winter times. Nothing like eating preserved boiled fruits when the fresh are annoyingly unavailable).
Each blend/jar/kompot would be uploaded as a new version to this model. Anima versions only. I have Illustrious versions of most of these, and might upload them in a separate dedicated model.
Наздраве! (literal translation sounds weird, hmm) Поздрави! (I guess)
Description
Nebulous understanding of the surrounding universe. On the microscale, the immediate surroundings, too. Must be nice, viewing wisdom as vice.
For the confused and hungry for the knowledge (in advance) of what this style is about, please refer to my reply in the pinned comment.
FAQ
Comments (12)
Could people start to actually describe the model in their descriptions please
All i learned from this one is "it's two art styles combined", like you could at least say what art styles? general image composition/subject?
@almostred Not with this author, bud. Sometimes, things are different than the usual - "name, follow me here, so I can earn money from you, join my social so I can hook you, so I can earn money from you, and don't forget to follow me here, to see my regularly scheduled uploads", increase the font, add some color to it and copy-paste it everywhere, on every model. Art style is a visual matter. Anima, the base for this LoRA, has a massive and perfectly working artist styles included, so one can argue LoRAs are not needed for artist styles. So, I have decided to share mixes that try to provide some benefit, as in add uniqueness without the need to add "@Artist1, (@Artist23:1.2)" etc. and look distinctly enough. General composition/subject? Say what about them? That is made by each user through the prompt. The LoRA is spice to the dish the user made. Sometimes a bit of intrigue adds to the memorability, adds the option for trial and discovery.
But some people prefer to be spoon-fed, I suppose. Betelgeuse is blend akin to Kunaboto's style. Nebula B is a massive mix of 5 different art styles with an emphasis on softer shading, smother colors, and subtle or nonexistent linework.
And yes, why don't these damn people actually describe their model in their descriptions.
Why don't these other people actually read the descriptions?
leave our autistic girl alone
i come here for descriptions 1st and models 2nd
As a guy who love mixing styles, this is a gift, lol.
I like this idea, interesting to see what other mix you gonna made. For me, it's somehow looks like devilHS + kunaboto, lol
Do you mix styles just by putting several styles in a dataset?
@SillyZilly well, yeah
@ElfSimp no
@SillyZilly I thought you were asking me, lol
@ElfSimp yeah
Such a great conversation!
I'm still unaware to whom the question was, but regarding style mixing, I usually follow several methods.
1) Single dataset - put everything there. Easiest to do, but weakest results. Resulting images tend to be either or depending on the step the model was saved on;
2) Different subsets with different repeats and dropouts - harder to set up, better results. Slight risk of underfitting if not trained properly;
3) Use AI, or your own hands, and make images that combine the targeted styles. Build a dataset from them.
4) Any sensible combination of the above options.
@PulenKompot maybe