File: 00140-4245825385-20251001170603.png (877.7 KB)

877.7 KB PNG

Request Anchor

>>11454604
Ouro Kronii with four breasts please.

>>11454604
Anything along these lines, or more from this if anyone has it?

>>11453625
>>11453626
>>11453629

>>11454604
Huge nipples
Big enough to fellate

>>11455404
Very nice!

File: 20260113233921.jpg (185.1 KB)

185.1 KB JPG

Good news! Despite selling out months ago, you managed to score an invite to a very popular fetish ball taking place all week!

Bad news... You didn't actually know the event was taking place when you booked this hotel, and your impromptu invite was not exactly... Optional.

>>11454961
Seconding this.
Possibly something a lot like the current first image in the /ddg/ thread with the horsecock nipples but without the futa part.

File: 00010-1834828814.jpg (399.6 KB)

399.6 KB JPG

>>11454961
>>11455815
This was as big as I could get without having to fuck around with finding a LORA for it. Hope they're girthy and fellatable enough for ya

File: 00013-2768637082.jpg (251.7 KB)

251.7 KB JPG

>>11455815
>>11454961
And one more with a different vibe

>>11454604
gimme some slim to hyper BE, easy. of you want extra credit, add growing an extra pair of tits too

>>11455995
>>11456159
Thanks very much!

File: ComfyUI_22607_.jpg (362.9 KB)

362.9 KB JPG

i liek big butts

>>11454604
requesting hildr from fgo with an amazonian build working out at the gym in a pink sports bikini while ortlinde and thrud watch in awe

>>11456722
>>11456723
>>11456724
>saved on /e/
>saved on /d/
hehehe

File: hentai3C.png (127 KB)

127 KB PNG

>>11454604
requesting this pic of the girl from genzo hitogata kiwa be colorized.
orange for her hair and pubes.
grey for the scythe blade.
light brown for the wood where it meets her arm and the rope near the bottom.
off white for the bandage that binds the wood and scythe to her arm.

alternately i would like pics of one armed woman with eyepatch, split hook prosthetic, and pegleg or any combination of those 3.

or even just a pic of an able bodied guy being raped by a completely limbless daruma woman. The woman must be completely naked and not using any prosthetics and it should be obvious the guy is not consenting, maybe because he's tied down. The guy should also be naked.

File: 00022-434398821.png (1.7 MB)

1.7 MB PNG

It's taken me a long time to figure out the Pair captcha.

>>11457093
More of a general question to the thread. Would it make sense to ask for a bucketfill-colored image instead of just lineart as reference? Not interested in this request, but if I were then that would be my first step in doing it, to prepare for img2img at partial denoise and line controlnet. Might as well ask the requester to do that, instead of listing out the colors in text. Or do you use a different approach?

File: 1961599177883662070.png (1.6 MB)

1.6 MB PNG

>>11457243
not sure on how to answer your question, but i'll give you an advice on bucketing an image for i2i- sd seems to get confused by large swatches of solid colour, adding some noise helps. my usual workflow is rough->manual redraw->i2i, i leave all the changes i did on a single layer and send that layer to paint.net's default "add noise" plugin before feeding it back to sd. a very low value, you should barely see the effect when zoomed out, but it's gotta be there when you zoom in. you can try ghetto-ing it with low quality jpeg, but paint.net is free.

>>11454604
Sportsball player taking his reward for winning the big game. (cockvoring some of the cheerleaders in the locker room)

>>11457344
I have not experienced that. After all it adds random noise into the image already as part of the img2img process, that's what denoise is for.

File: 00125-2881210372-20260114150349.png (1.2 MB)

1.2 MB PNG

>>11456722
>>11456723
>>11456724
Do you have a catbox for any of these? I love the style!

File: hawto.png (1.1 MB)

1.1 MB PNG

Ah, crap. Didn't realize there was one of these up. Postan the one I posted there + a new one.

File: iyg.png (201.8 KB)

201.8 KB PNG

>>11457472
i've had it leave the bucketed area pretty much unchanged, as if it added noise, saw that there's barely any in the solid part of the image, and said "yeah, this should be solid".
this is a fail pass of >>11455404 where i forgot to add the noise and instead of blending it in it just kept the blob of solid colour, while darkening the ceiling around it so it outright blended in worse than in the fixed rough gen it got on the input.
but maybe i'm just doing it wrong. i definitely do it unorthodox.

>>11457228
>may without her boobas
Trash

>>11454604
Pregloli seggs of Kanna, Lana, Becky, or any loli of your choice. Catbox it obviously

File: thingiething.png (975.2 KB)

975.2 KB PNG

>>11457093
I did things my way.

>>11458109
this is wonderful! many thanks!

>>11454636
>>11455513
>>11455620
>>11457344
How do you achieve this level of amputation?
I'm using Quadruple_amputee/4_missing_limbs but it never manages to fully remove the whole limb.

File: 1963675557442563355-1.png (1.2 MB)

1.2 MB PNG

>>11458738
gotta need loras for that, but they aren't hard to find on civitai.

>>11459762
Posting here since slower traffic than the other

>I like the textured line art look and the slight noise covering the whole image.
HSV Noise at 30% opacity lmao

>I think one thing I've come to terms with is that depending on the checkpoint and lora, I can't just shoehorn something that the AI doesn't understand or something that is uncommon.

Some models are more flexible than others, but it's also a matter of finding the right tags. A good trick is to go to danbooru to find uploads of something similar you're trying to make and see how it's tagged.

>I can paint digitally quite well, it's just that when the AI produces images that has a shit ton of gradients and smooth shading, it becomes a bit of a headache to make changes.

You undertand why I keep it flat and simple now. I'm good at imitating a variety of shading styles too, but it's not always easy or quick to keep it seamless and perfect, and I don't feel people appreciate smooth stuff as much now that AI can do it in a pinch. Matter of taste at the end of the day

>It's a general question, but do you have recommendations for any LoRa that compliments a checkpoint? I use some Illustrious models at the moment - I like the look but sometimes the faces gets too anime for my liking.

Lora behavior is sometimes unique to the checkpoint and whatever you're aiming for. I recommend you hunt for a good general purpose checkpoint, minding CFG when testing. Make a test gen and if the colors are too saturated lower it, increase it if it's too gray. The colors in question don't good when CFG is off center, high cfg isn't vibrant but looks burnt in, likewise bad for low cfg, so I wouldn't recommend using cfg for color control.

>>11459905
>>11459762
>Another more specific question: Should one always use a LoRa that is trained on a specific model or can they overlap - For example; can a LoRa trained on Illustrious work with a Pony model?

It can and it will but not well. The way you can mentally visually it is DNA and checkpoint lineage (without inbreeding), the closer the checkpoint's data is to the lora's checkpoint (that which it was trained on) the closer the intended results.
Pony is a stranger to Illustrious, while NoobAI descends directly from illustrious: An illustrious Lora will then work excellent on illustrious, OK on NoobAI, and badly on Pony.
LoRas are sometimes made for specific merges, or may be epred/vpred. Just mind what the site says and try to stick to what you have. If pressed because no other good loras are available you can try, it can sometimes surprise you and work better than expected.

>I am asking because sometimes I can't quite tell if it's working or not.

When testing loras for the first time, keep the same seed and generate with and without lora. This will tell you if it works and also to what strength it changes things. Don't test with controlnet or i2i of course, as clean a t2i gen as possible.

>>11459906
>>11459762
As for controlling style, models...
https://danbooru.donmai.us/wiki_pages/tag_group%3Aimage_composition
Good tags for clean shading:
flat color, anime coloring, oekaki, cel shading, sketch*, lineart*
*sketch and lineart can be used on a colored image if used after the colors set in, if you do something like: [:sketch:0.4] <- similar to controlnet endpoint, apply after 40% is done

3d and realistic on negative can also flatten the color

Artist tags in practice rarely copy the artist's style well. But you can use them to influence things like shading and line weight. A caveat is that you don't get what the artist drew, you get what that means for the checkpoint, so always isolate and test, an artist tag can be useful even if the artist isn't what you're going for and vice versa.

Faces may be the exception, artist tags do influence that a lot. So if you don't want a typical anime face as you mentioned you can try complementing with some western, western-anime artist style types.


There's no single recipe for this, you gotta brew your own stuff, but once you get familiar with it you'll know exactly what to tweak to get what you want in a pinch.
For models, I'd encourage you to at least try the most popular ones. Go generic WAI first, WAI is OK
Just remember: a checkpoint is not how it looks at base style, it encompasses how all tags interact. In fact, checkpoints with strong base style tend to get in the way of your style tags. You can make a checkpoint look 180 to its base style with tags if it is flexible enough.

LoRas: use them to compliment. circustantially if you need a rare character or odd concept. As general tweaks: keyword is "slider", but always test them, a bad LoRa can affect style or downright poison your gens. Test with, test without, greenlight or delete.

Good luck

>>11459905
>>11459906
>>11459907
Eating this up rn. Thanks for sharing and explaining so thoroughly - I've been experimenting a lot, especially with just getting used to the knobs and sliders and what to do and what to avoid. Like you said, I gotta brew my own shit. But I can already see a lot of potential with producing a lot of things quickly.
I am also having fun, so there is that too.

>>11459937
No problem. One last thing I'd mention is, if your go to method is sketching, then don't factor in body proportion when balancing your style and artist tags. Artist tags can influence things like head to body ratio heavily. But if you're already guiding that, then focus on what affects shading, linework and color the most.

And remember you can weigh all tags with (tag:value) -> such as (pastel colors:0.8)
or control endpoint with [tag1:tag2:value] -> where for [:pastel colors:0.5] would mean tag1 is empty tag2 is applied on the latter half of the gen and [pastel colors::0.5] would mean pastel colors is applied on the first half

The last thing is not as useful when using a sketch source since you're not starting from scratch and the base is already set, but it's useful to know anyhow. The more tools available the better, sometimes you'll find yourself using different methods to do different things and full control over the denoise process comes in handy.

And yeah, do have fun. This shit can be very enjoyable to fuck with when you already know how to draw, it's a lazy power multiplier.

>>11459973
Thanks for sharing.
I do find that sketch with img2img and control net has been on average the best for what I want right now. I do change up my tactics depending on the situation - Playing with sliders and such, changing tags, playing with control net, etc. Right now (at least for me) the Control Net Line art combined with Reference is pretty solid - It doesn't deviate from the colors much at all when I do that - and depending on the setting on reference I pretty much tell it to "don't deviate from this too much" and it's been pretty successful.
What has not been so successful is if I already have a pretty rendered piece already but it's somewhere in the middle and some parts are clean while others are not it actually struggles a lot more - So i'll stick to sketch ->>> Ai ->>> final touchups in Photoshop.

Also regarding this >it's a lazy power multiplier.
One hundred percent. I have already finished some personal goon art that I wouldn't have bothered to do because I'd rather do other stuff, but it really is a power.

>>11460112
This is the sketch used. I used img2img with Lineart Control Net and Reference Control Net. Timestep range for both was around 0.55 end. I was using WAI illustrious only, but then I fiddled with some artist lora and ended up changing plans midway.

>>11460112
>>11460113
>>11460114
good job, It's such a close detail sketch you must barely need controlnet. probably good to have for reinforcement.
Clean, nice.

One detail though: if you're gonna do furry, anthro, monsters, animals you might want to choose a noobai (or pony) checkpoint merge.
Illustrious has danbooru data only, noobAI combines illustrious tags with e621 stuff (the big furry site). WAI is illustrious based

This does not only include furry stuff but a lot of /d/ related content. I'd recommend noob over pony since noob will be compatible with all your illustrious stuff and pony is a bit dated, having other downsides such as obfuscated artist tags you can't use.
Use whatever works for you though

>>11460128
Yeah I suppose it was a bit overdone with the sketch, but it helps a lot - I basically want it to render for me. I guess I'll try even more simple sketches if I feel lazy.
Regarding the WAI and furry - Yeah I am aware. In fact I have been stress testing this model a lot with very non anime or ugly looking stuff or even stuff it wouldn't consider, just to see what I can get away with. I will in the future get Noob ai though, thanks for the tip :)

>>11460128
Also on a random note - Can dropping LoRas in your LoRa folder potentially cause conflicts via accidental trigger words or does this not matter as much?

>>11460138
Loras won't be active unless you put them in the prompt, showing as

So no

>>11460137
Do try, and be aware there are many different e621 tags not recognized by the danbooru autocomplete plugin, if you're using that. noob and pony will recognize those, so you may want to fetch them from the site if you need specific stuff

File: 20260122045702.jpg (264.6 KB)

264.6 KB JPG

Had a dream about this somewhat symbiotic alien insect, that will crawl head first into the anus and consume any waste it finds there, laying its body along the person's spine and wrapping its legs around their torso. The bug itself creates very little waste, eliminating the need for complex waste disposal, and being from space, seems to confer some small amount of protection against cosmic radiation, so 'wearing' one of these bugs becomes mandatory for long term spaceflight, like service aboard a starship. Uniforms and space suits are designed around it.

But upon first entering, it creates a knee buckling orgasm and long term exposure could have some mind altering consequences but I woke up before I found out what it was.

https://files.catbox.moe/o2wf1m.mp4

i'm actually slightly proud of myself for this one.
mostly cuz i managed to get the slime egg laying to work without a lora.

File: 00014-3898955652-20260121143713.png (690.4 KB)

690.4 KB PNG

>>11457678
Always love your stuff, this Asuna is great!
Anyway, here's a chick whose people-eating habit has invalidated her new year's resolution to lose weight.

i'd been following method 2 in the nividia gpu guide but keep getting this error when trying to run webui.bat looking through the other guides don't know what i've done wrong

"exit code: 1

stderr:
D:\python.exe: No module named pip

Launch unsuccessful. Exiting." i got python installed too so i'm stumped

>>11461252
The old automatic1111 WebUI hasn't been updated in more than 2 years. Although in theory it "should" still work, it may be hard to get tech support. Try Forge instead.

>>11454604
The Tyrant from "Final Fantasy XIV" hooked up to a lactation machine, chained to a bed, with a bunch of already-made bottles of milk, and he's currently milked, making an ahegao face.
(passing on a request)

File: 00048-24160566.png (1.8 MB)

1.8 MB PNG

>>11461344
thanks

File: 00050-3707955996.png (1.9 MB)

1.9 MB PNG

>>11461973

File: 20260125090105.jpg (150.6 KB)

150.6 KB JPG

>tfw no latex gf to watch the end of the world with

>>11462095
no end of the world to watch, either

latex is not a personality trait though, plenty of girlfriends could probably handle that as a one-off

File: Misty_Vaporeon.png (1.5 MB)

1.5 MB PNG

>>11454604
Requesting HGSS Misty as a Vaporeon siren, having two flexible Vaporeon tails for legs and doing the pose as in pic related. Two versions, one nude, the other one with her swimsuit.

Can I get Yuri from DDLC as an amputee nugget laying in bed naked wanting you to fuck her?

File: final true real.png (1.4 MB)

1.4 MB PNG

Tummy listening is so intimate...

noob question:

i'm trying to gen 2 characters but want to get specific on each, purpose is to not have details bleed from one to the other.

is inpainting the only solution?
prompt magic isn't really consistent.

>>11465028
regional prompter

>>11465028
>>11465029
https://rentry.org/sdgguide#regional-prompter

>>11465029
>>11465036

much appreciated, jumping in the rabbit hole now