following prompts: Anime art full character concept art, high definition anime art, finely detailed, adult anime girl, violet evergarden, platinum blonde, walking, scenic, highly detailed, sun, grass, trending pixiv, landscape scenic view of baroque cathederal, European, illustrated by Makoto Shinkai, trending on pixiv fanbox, oak trees , extremely high quality artwork,
why do the traddites loathe and fear what they don't understand? Anxiety turned to violent action, I just can't understand it. We have given them everything, and they repay it in animus.
>github.com/hlky/sd-enable-textual-inversion so is there a particular reason to use a placeholder name like *? or if I am training on pictures of Goofy, can I just train it on "Goofy" and then import it and suddenly my SD can actually draw Goofy instead of these weird half-dog half-mouse things?
Christopher Perez
anyone know what this specific type of veil or whatever is called? I can't seem to get the AI to replicate one I've tried stuff like hijab, arab veil, belly dancer veil, face veil
"*" is only used as it guarantees to only take 1 token, which you need for this. You could also use @ or # or short words like "his" "the", as long as it is one token only
Lincoln Perry
The model already associates "goofy" with other shit.
You people are not "artists". Putting a bunch of words into a machine and getting it to generate shit for you isn't "art" by any stretch. You have no artistic talent. You likely have no imagination either. You all use the same fucking prompt keywords when you use this shit. And you shouldn't feel proud when you post AI Generated Trash #195,172, because you've accomplished nothing. It's sad as fuck watching you clap each other on the back for shit that you didn't even do. When I finish a piece, it's 100% my own creation that I poured my soul into. You'll never feel that sense accomplishment, to see your imagination, skill and talent work hand in hand to create something original.
Jonathan Harris
are you using just the character name or the actors name as well? this prompt kept giving me shitty hugh lauries until i put hugh laurie in the prompt, it also somehow improved the lisa cuddys without me adding the actresses name in gfpgan ate the faces a little
They're looking at it too, it is the next step. Haven't seen anything concrete yet in this direction though, besides some example images and pinky-swears.
textual inversion is NOT training or fine tuning your existing SD model.
Justin Ramirez
you can use any single token character, rinongal one of the authors of textual inversion explains it better github.com/rinongal/textual_inversion/issues/14 check the history of the readme in the repo for my thoughts on a way to do merged embeddings and have better prompt words, the end result should be that you would be able to merge any embeddings you come across with your own embedding file it would assign a single token placeholder, a placeholder to use in the prompt that's user friendly but still unique, and then a name for it that would be shown in a list of the embeddings in your merged embedding file so you know what it is
Oliver Gutierrez
I give it maybe 5 years before AI puts you out of the job. How's that feel?
Lincoln Rogers
I am an artist and I use AI as a tool. And no you will not see my work in these threads, though I like seeing other artists incorporate this in their workflow.
>open vino finally, I've been spamming this crap for a week now here and on aids, they finally added openvino to tuttards, I wonder how many people know that this crap supports not only intel and amd but also ARM including Raspi and Android (although plug-in support sucks as of today)
Charles Morris
>No >Rules >Only >Tools
Ian Wood
though I will add 99% of this shit is garbage y'all having fun though, nothing wrong with that. The difference between my dad using this and me using this is huge
Brayden James
I'm an artist and I think its great that there's a tool non-artists can use to make cool images. Grow up.