/hdg/ - Stable Diffusion

Anime girls generated with AI

Index Catalog Archive Bottom Refresh
Name
Options
Subject
Message

Max message length: 12000

files

Max file size: 15.00 MB

Total max file size: 50.00 MB

Max files: 4

Supported file types: GIF, JPG, PNG, WebM, OGG, and more

E-mail
Password

(used to delete files and posts)

Misc

Remember to follow the Rules

The backup domains are located at 8chan.se and 8chan.cc. TOR access can be found here, or you can access the TOR portal from the clearnet at Redchannit 3.0.

Uncommon Time Winter Stream

Interboard /christmas/ Event has Begun!
Come celebrate Christmas with us here


8chan.moe is a hobby project with no affiliation whatsoever to the administration of any other "8chan" site, past or present.

>>25869 Yeah I don't even think I've ever seen a realistic goblin anywhere ever. I'd go as far as to say it's out of the realm of what Stable Diffusion can do (right now) If you're really committed you could try what >>25866 suggested, but from experience i2i without base data tends to murder quality. For example, if it doesn't know what a big goblin ear is, it will find the closest thing resembling to it, as for everything else, it'll become more human with each pass. Controlnet is also an option I guess, although I'm unsure what particular one would be best for that project.
>>25872 GHOST CUM DICK GHOST CUM DICK GHOST CUM DICK
>>25871 Have you tried everclear and this https://civitai.com/models/428826/damn-ponyxl-realistic-model Last one seems to have some fantasy concepts. Maybe it can handle a short green bitch with pointy ears?
>>25874 Holy shit that model is CRACKED lol'ing at the variety of degeneracy and brainrot, and it looks pretty consistently good. Like maybe one or two gens here and there have some odd finger additions but nothing extreme. I'm gonna give it a shot. >that f*male gator
>>25875 Yeah that's why I suggested it. You could also probably do an image search for realistic gobbos and then check the resources used. Can't check rn, but there's a chance
>>25876 DAMN wins so far, it got the idea first try and the only problem i'm having is really grainy hands and feet, but the anatomy isn't wonky and she looks very goblin. prompt strengths are overly strong, prompting a big belly and tits gives her SSBBWW or some crazy fetishy shit.
>>25876 >>25868 uploaded some cherrypicked of today's fails, apparently i have almost 300 of them. theyre noteworthy for lora mixing/no lora just rawdogging a model's capabilities/trying negs (i think?) and crazy prompt experimentation this is certainly a nice combo of challenging gens to try for https://catbox.moe/c/n7mrij
damn that's a lot of posts for this place
>>25872 Spoilered image is the best one.
>>25879 mostly me
>>25881 tan brats in white clothing flashing pink nipples ToT
i think i found a new autistic fixation, trying random schizoid civitai prompts and testing new model prompt adherence through it very interesting stuff.. >view from side, booty, wet and dripping with ky jelly, body completely drenched and dripping with honey and liquids, sitting on the floor with legs wide open, small chest, dynamic posing, moaning, ultra-detailed facial features, burkinabe woman, sexy plump bbw thick curvy plussize light skinned black woman, proportionate bbw body, cute, hyperdetailed, huge ass, wide hips, stretch marks on hips, perfect big butt, plus size ass, thick chubby legs, full makeup, naked, being slutty dripping thick honey, completely drenched in honey, completely covered in honey, sultry expression, seductive, hyperdetailed, 32k uhd, extra sharp focus and details, 100 megapixel highly detailed,
>>25880 because that's the specific drawing that it's aping, lol
>>25883 they are still the cutest in micro bikinis
>>25886 funniest, yes. They're cutest in oversized clothing that has too many layers and frustrates them because they can't seduce ojii-sans
>>25888 peak SEXO
@BJD anon I'll bake you a LORA if you can give me dataset material
>>25890 sure, itll be a while though. Finding consistently decent pictures of these things is surprisingly tough. I gave up when i realized i maybe had like 5 good images in my dataset kek. Lotta watermarks/censor emojis/124px compressed jpegs/etc..
swear on me mudda i'm never gonna fuck with negs in any XL/pony model, if a LORA suggests negatives, its a shit LORA. simple as.
(2.40 MB 1664x2288 catbox_kilcxz.png)

(2.77 MB 1664x2288 catbox_z8pc79.png)

(3.30 MB 1664x2288 catbox_sgdi75.png)

(3.05 MB 1664x2288 catbox_0tlsaq.png)

>>25888 box for rin?
>>25813 got my 3090 in and it does about 2.6 it/s on the same settings definitely more card than i needed in retrospect but i'm happy to have it
>>25896 >3090 download mixtral instruct 3.7bpw exl2 and enjoy slippery sloping down the LLM rabbit hole
>>25896 and more importantly, it can do 1 it/s while i'm playing video games >>25897 i'm looking forward to seeing what local models can do, claude opus was fun but sonnet doesn't have the horsepower for what i wanted
>>25898 mole
is image interrogation meant to be insanely shit? >takes forever to load the model >finally loads >cant caption to save its life tried two models and had to reload both times because neither could unload properly..
>>25901 its the "ffs I missclicked that button again time to wait a minute" button
>almost 1200 someone make a new thread soon please thanks
(29.59 KB 474x315 bread.jpg)

>>25903 Bread baked: >>25904 >>25904
>>25900 Box for the tianliang-looking one pls


Forms
Delete
Report
Quick Reply