#Solo Leveling Artstyle TI
77 messages · Page 1 of 1 (latest)
do it
im doing an experimental LORA now, no promises its going to be anything good
niceeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee
thats dope, but is it train on jinwoo aswell
I mean you could get fairly close if you just prompt his characteristics
but I assume you could also use the jinwoo lora on civitai with this and it should work
Primarily it was made as an experiment, and was using a custom merge with anything4.0
this is the custom merge
this is the same prompt with anything4.0
again, with prompting it should be fine enough
however you may want it
when did this drop
beautiful
just doing some tests before I upload
how many images did u train
think it was like 76
I mean, the style was pretty consistent
honestly getting better as it went
@hexed shuttle whats your preference. unfortunately I trained it on a custom merge. Should I just upload the model and explain the merge, or just upload the merge as well
thats so goooood
i want the purple aura-ish look and the black suit
getting the thin line aura might be a little hard
lora
if you use it with that sung jinwoo lora you should be fine
it seems to have the outline
but is also incredibly focused on one look lol
actually it shouldnt be too hard to get the glow actually, just needed to prompt with 'purple glowing outline'
aight ill get uploading
damn why did the style change tho
are u making 2 diff lora's? i dont understand
i thought ur just further training that one
someone already had made this one
just for sung jinwoo
Pleased to present a Solo Levelling style LoRA. Wasnt trained with any NSFW in the dataset, but by all means it can be NSFW if you prompt for it because it was trained on a custom merge which included Anything v4.0.You may find this merge here if you do not wish to merge it yourself (Details below).You can weight it from 0.7-1 and you should get...
nop
how do u even train a already made lora
bruh I just completely made my own
someone else also just happened to make their own, but specifically for sunjinwoo (in that specific form)
yes i know i got that. im just wondering how to train a lora further 😭
oh
well it would require you to have the authors dataset and training info
when the author does a v2, they typically just do an entirely new train, but add some images and tweak some settings to make a better version
yeah, i trained it on everything
i mean it wont make fight panels naturally, but it can still make some cool ones following the style at least
you could make do a background panel, then inpaint some characters (or perhaps using controlnet)
lol
cant get a proper result
what model should i use? im using v3
anything^
It says on the page
It was trained on a merge of Anything v4.0 + Offset True darkness
**Anything v4.0 **should get you some closer results, but ultimately if you want what I genned you need my merge (Which I also uploaded)
Its a bit of a pain, but I just did this train as an experiment so didnt really set it up to be sharable
can u send me the dataset if u still have it
my images and tags?
yeah just images