[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]

/vt/ - Virtual Youtubers

Search:


View post   

>> No.91952063 [View]
File: 3.02 MB, 1328x1944, 20259-20241221060711-2440756370-28-1328x1944.png [View same] [iqdb] [saucenao] [google]
91952063

>91806005
Loras I'm probing/interested in/sitting on
Koganei Niko - Been putting this off a lot, but I'm also waiting for a decent amount of fan art that accurately depicts the jacket, lots of artists are not even bothering
Perry Winkleton - about 30 images, doable but limited because mostly model shots
Goro Gaomi - also about 30? same deal
Hysteria.EVE - around 20 images, looks like a nightmare because of so many little bits but still technically doable
Vee-sensei - 25 images? Simple design so should be pretty doable, but almost entirely model shots
Kitanya - 30+ images for old and new, all it really needs is the work. Would also need quite some time since I'd rather they be in one lora and I dread any possible conflict
Nova Aokami - Plenty of art but the main issue with this one is what seems to be 3 different outfits for someone that doesn't get a ton of art. Even if I were to only aim for the first outfit that cuts out a chunk of the dataset
Dr. Nova(e) - My last lora for her had spotty recognition of her thighhighs so the dataset needs to be revitalized (with some extra leg-included images - artists seem avoid them for some reason) and restructured (getting rid of jacket overrepresentation and other redundancy)
Kiri Kilovolt - Mostly model/reference art but plenty of images. The main issue is outfit is mindfucking the network, even just getting to a decent place with everything on + jacket off. Stickers are also a nightmare
Riki Poppet - Doesn't even seem to be worth pursuing right now
Shiki Miyoshino (husky) - 30 images or so. Seems doable, but I also want to include everything in one lora, which makes it a lot
(questionable number of images is due to un-split character sheets)

There's also the tons of loras that haven't been ported from Pony... Thankfully it seems illustrious loras work pretty well for illustrious derivatives, though I haven't been messing with the vpred derivatives yet. I expect to very slowly work through them until I get a gpu that can train SDXL loras in SD1.5 timeframes. If I really wanted to I could write a script that handles a queue and everything that I could run whenever I'm not planning on generating smut, but aaaah

Have a hime and some quick mozu shit I will jack off to later
https://files.catbox.moe/pkemtv.png
https://files.catbox.moe/rxyd2o.png
https://files.catbox.moe/o8e26p.png
https://files.catbox.moe/lvqeka.png
https://files.catbox.moe/6i69th.png

Navigation
View posts[+24][+48][+96]