Sorry I haven't done one of these in a while but I'm back to share some more artist tag combinations that I've been playing around with lately. This is also a sort of preview of the next version of Duchesse as that's the model I made these images with. I wanted a model that could more faithfully reproduce artist styles. It will be published once the current Duchesse bounty is over (I really made that thing waaaay too long).
These are a mix of tags I was trying out, and some combinations taken from prompts I saw on the site. I noticed that there are a lot of incase style LoRAs so I tested a bunch of combinations with that tag and included an image with just the incase tag so people can judge what the tags can do without any LoRAs.
This is the base image with no artist prompts. The cover image is the same prompt but with a different aspect ratio and these additional tags: realistic, real skin, impressionism,
Base prompt: masterpiece, best quality, amazing quality, very aesthetic, absurdres, newest, scenery, anime coloring, flat color, cowboy shot,1girl, solo, orange hair, short hair, thick eyebrows, hair behind ear, detailed eyes, light green eyes, nose blush, looking to the side, looking to the right, dark green sweater, ribbed sweater, gold buttons, beret, indoors, cafe, green wall, large window, window sill, window frame, depth of field, blurry background, breakfast, coffee,head rest, sitting, table, smile, sidelighting, sun rays, shadow
moyazou \(kitaguni moyashi seizoujo\), (haimura kiyotaka:mimonel:0.9), hong_(white_spider)
incase
(onikobe rin:mossacannibalis:0.9), (asteroid ill:chigusa minori:0.7), (bigrbear:dandon fuga:0.4) ,(morino hon:eroe:0.85)
(koyorin:chengongzi123:0.7), nishieda, (koto_inari:koyorin:0.5), (shirabi:kairunoburogu:0.7),(kagami hirotaka:lasterk:0.7),
(incase:chengongzi123:0.7) (kotteri:masha:0.4) ( himukai yuuji:1.2)
artist:john_kafka artist:nixeu artist:quasarcake
(raichiyo33:sam yang:0.6), (barleyshake:shiontaso:0.4)
(incase:chengongzi123:0.7), (kotteri masha himukai yuuji:1.2), [meyoco:optionaltypo:7]
I removed the tags "scenery, anime coloring, flat color" from the prompts for these images
(alkemanubis:khyle:0.3), guweiz, (z3zz4:sakimichan:0.6), (opossumachine:incase:0.4)
koto_inari, (pentagon \(railgun ky1206\):0.6)
(alkemanubis:khyle:0.7), (z3zz4:sakimichan:0.6), (opossumachine:incase:0.4)
saonserey, Guweiz, Riz, Noriuma
Edit: I should also have mentioned this; I have found that some checkpoints can handle multiple artist tags better than others. I'm not just talking about the style but whether or not you can get a coherent image at all. Some models seem to occasionally output either black images, or just noise depending on what tags are used. So if you try these tags and you get some garbage results try switching to a different checkpoint or removing some of the tags.
Earlier Articles: