[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/aco/ - Adult Cartoons


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: asdg.jpg (636 KB, 2048x1024)
636 KB
636 KB JPG
Previously: >>8295085
The cartoony side: >>8304953

No DALL-E stuff.

>Guides
General: github.com/AUTOMATIC1111/stable-diffusion-webui/wiki | rentry.org/sdg-link | rentry.org/rentrysd
Hires Fix: rentry.org/hiresfixjan23
Inpainting/Outpainting: rentry.org/drfar | rentry.org/inpainting-guide-SD
Animation: rentry.org/AnimAnon
Create LoRAs, TIs: rentry.org/59xed3 | rentry.org/simplified-embed-training
List of models/LoRAs:
rentry.org/acorepo

>Auto1111 extensions
ControlNet: github.com/Mikubill/sd-webui-controlnet | rentry.org/dummycontrolnet
Regional Prompter: github.com/hako-mikan/sd-webui-regional-prompter
Tag autocomplete: github.com/DominikDoom/a1111-sd-webui-tagcomplete

>Related
>>>/h/hdg
>>>/e/edg
>>>/d/ddg
>>>/g/sdg
>>>/trash/sdg
>>>/vg/aids
>>>/vt/vtai
>>
File: zonkeybondage1.jpg (239 KB, 768x1344)
239 KB
239 KB JPG
Slowly getting into zonkey/pony prompting
>>
>>8315060
Make some hairy Sadie Sink please.
>>
>>
File: zonkeybondage2.jpg (252 KB, 768x1344)
252 KB
252 KB JPG
>>8315074
>>
File: zonkeybondage3.jpg (252 KB, 768x1344)
252 KB
252 KB JPG
>>8315124
>>
File: zonkeybondage4.jpg (266 KB, 768x1344)
266 KB
266 KB JPG
>>8315128
>>
File: zonkeybondage5.jpg (295 KB, 768x1344)
295 KB
295 KB JPG
>>8315134
>>
File: 75637y567.png (1.56 MB, 1024x1024)
1.56 MB
1.56 MB PNG
>>
File: 6456747.jpg (2.07 MB, 2048x2048)
2.07 MB
2.07 MB JPG
>>
File: zonkeybondage6.jpg (308 KB, 768x1344)
308 KB
308 KB JPG
>>8315138
>>
File: zonkeybondage7.jpg (328 KB, 768x1344)
328 KB
328 KB JPG
>>8315230
>>
File: ComfyUI_00086_.jpg (1.27 MB, 2048x2048)
1.27 MB
1.27 MB JPG
>>
>>8315138
Where do I get Zonkey?
>>
File: zonkeybondage8.jpg (306 KB, 768x1344)
306 KB
306 KB JPG
>>8315236

>>8315408
on civitai: Zonkey Realism
there you also find images at the bottom of the page, most of which show the prompts and settings used
zonkey needs a bit of setup and is a bit more vram hungry than other checkpoints
>>
File: zonkeybondage9.jpg (307 KB, 768x1344)
307 KB
307 KB JPG
>>8315419
>>
File: zonkeybondage10.jpg (280 KB, 768x1344)
280 KB
280 KB JPG
>>8315421
>>
File: zonkeybondage11.jpg (311 KB, 768x1344)
311 KB
311 KB JPG
>>8315425
>>
File: zonkeybondage12.jpg (322 KB, 768x1344)
322 KB
322 KB JPG
>>8315431
>>
File: 00066-1900119004.png (1.65 MB, 1024x1024)
1.65 MB
1.65 MB PNG
>>
>>8315419
Can't find it. Might have been removed. Shit. Awesome pics btw.
>>
File: 21093-108115505.jpg (511 KB, 1344x1728)
511 KB
511 KB JPG
>>8315834
https://civitai.com/models/405136/zonkey-.realism
(pic not related)
>>
>>8315876
FUCK YEAH BROTHER! I have tried everything in the search. I even wanted to type Z (kek) and go down the list. Nice. Thanks.
>>
File: 6454.png (1.29 MB, 1024x1024)
1.29 MB
1.29 MB PNG
>>
File: 0006.jpg (1.47 MB, 1736x1736)
1.47 MB
1.47 MB JPG
did the dalle3-lora emma watson poster ever make more like these?
>>
File: 0011.jpg (1.45 MB, 1736x1736)
1.45 MB
1.45 MB JPG
>>
>>8315928
I did the lora but not these images. Pretty great to see tho. new version going up late tomorrow night under a secret name that wont get taken down after 100 downloads
>>
>>8316081
did you put up sexual images for the lora on civitai?
>>
File: 00001-2819380057.jpg (1.57 MB, 2000x3200)
1.57 MB
1.57 MB JPG
>>
File: 65867.png (1.72 MB, 1024x1024)
1.72 MB
1.72 MB PNG
>>
>>8315431
>>8315074
>>8315124

catbox for these three? I keep trying to get the girls off the ground (tied arms up, dangling feet) but I just can't get there
>>
>>8316408
The feet up is basically "legs up,feet tied to ceiling,shot from below" and some luck. Most gens were not like that and I wasn't going for it anyways.
Case in point since the other two have the same exact prompt. Also "arms tied behind back" or "arms tied".
DPM++2M SDE, CLIP skip 2 and 30 steps.
Negative "score_6, score_5, score_4, pony, censored, furry, child" and the positive starts with "score_9, score_8_up, score_7_up, cinematic lighting, realistic, hyperdetailed masterpiece,1girl blonde woman"
catbox sometime later
>>
File: file.png (1.91 MB, 832x960)
1.91 MB
1.91 MB PNG
>>
File: file.png (1.74 MB, 832x960)
1.74 MB
1.74 MB PNG
>>
File: file.png (2.99 MB, 1280x1280)
2.99 MB
2.99 MB PNG
>>
>>8316551
and CFG is 4
>>
File: 00010-566450591.jpg (606 KB, 3000x4000)
606 KB
606 KB JPG
>>
File: 00024-4019212506.jpg (659 KB, 3000x4000)
659 KB
659 KB JPG
>>
File: 00039-2556846194.jpg (546 KB, 3000x4000)
546 KB
546 KB JPG
>>
File: 00046-2703877955.jpg (466 KB, 3000x4000)
466 KB
466 KB JPG
>>
File: 00051-1616292854.jpg (619 KB, 3000x4000)
619 KB
619 KB JPG
>>
File: 00056-3592030169.jpg (484 KB, 3000x4000)
484 KB
484 KB JPG
>>
>>8316081
Thanks!
waiting patiently for the loras emma anon.
Any chance for the old ones in a mega while we wait for new?
>>
Am I the only one who, after trying realistic pony models for a while, can't go back to 1.5 or xl models?

It is so easy to create scenes as you want with the pony models, just by putting the tags the model recognizes it perfectly.

And that's why it makes me so disgusted to have to go back to 1.5 or xl models, because of how difficult it is to create scenes with their very long prompts, having to use several loras to create a scene or sexual position, that it's impossible for me to think about going back to 1.5 or xl models.

The realistic pony models are not at the same level as the XL or 1.5, in that we agree, with much more time and effort you get a better quality in XL or 1.5, but taking into account that only 3-4 months ago that have become popular I think it is advancing at a giant step.

I still think that the realistic pony models lack a bit of quality in the lighting and the details of the skin, I understand that it is because they are not designed for realism and are very trained in anime, but I prefer that, and wait for it to improve, than having to relearn the thousand things needed to get a good photo in 1.5 and xl.

Still, I don't know if there is any project that is entering a realistic XL model with the tags like pony does, because for example, I would like to use the face id plus v2, but at the moment the lora is not compatible with pony, a juggernaut/dreamshaper xL with the same style of creating images of pony (using tags, without so many loras), would be the best.

What do you think?
>>
>>8316719
Use whatever makes your life easier/gives what you want, everybody and everthing else can go pound sand.
>>
File: Saffron_Burrows,_006.jpg (29 KB, 512x768)
29 KB
29 KB JPG
>>
File: download (49).jpg (40 KB, 512x768)
40 KB
40 KB JPG
bumping for moar jenna
>>
File: 1717632165403273.jpg (509 KB, 960x1088)
509 KB
509 KB JPG
>>
File: 1704466793285059.png (1.05 MB, 680x1080)
1.05 MB
1.05 MB PNG
>>
What samplers/prompts are u using with zonkey?
>>
File: image (31).png (2.55 MB, 2432x1664)
2.55 MB
2.55 MB PNG
>>8316719
i'm torn, because i vastly prefer the aesthetic of the realistic models and the added eroticism of me being able to add my fetishes doesn't counterbalance my preference for the realism. Zonkey being an absolute pain in the ass with OOMs doesn't help matters.

>>8317087
just pull from civit the real open question is how best to upscale
>>
File: image (29).png (3.15 MB, 1248x1824)
3.15 MB
3.15 MB PNG
there's a tendency to think that if your gen was just hi-def enough, or had some extra little bit of je ne sais quoi, you would spontaneously nut, and you know what, that's not too far off
>>
File: 00061-474819662.png (1.41 MB, 1216x832)
1.41 MB
1.41 MB PNG
but like, yeah, congratulations, you made some coom shit. what next? I feel like with pony, the only way forward is degeneracy, whereas at least with realisic models, you can do some other shit with them
>>
>>8316191
No they were face closeups. Im not sure if some cuck reported me or what because the only way they could be sexual is via prompting. They were essentially a character/style lora

>>8316716
I just dont have pc access rn, the first one going up will be my latest.

Also does anyone have like 20 good dalle gens of emilia clarke at different angles and lightings/ eyes open and closed? Looking to train one for her
>>
>>8317177
Start from pony and merge/finetune.
If you want the max SDXL realism quality, you can use refiner at .8 with a epicrealism level model, but have your base model be ponyrealism. Ive used controlnet softedge to influence the setting and such, i.e. emma in the trunk of a car worked much better with controlnet
>>
After several months using the realistic pony models and seeing that one of my biggest complaints was that the images came out with too much brightness and contrast, I just realized that it is quite solved by lowering the CFG scale (I was used to use 7 in 1.5).

xd
>>
File: 00002-518281915.png (2.06 MB, 1248x1608)
2.06 MB
2.06 MB PNG
>>
File: 00001-1783203140.png (1.95 MB, 1248x1608)
1.95 MB
1.95 MB PNG
>>
File: 01163-3716199011.png (1.18 MB, 768x1280)
1.18 MB
1.18 MB PNG
>>
File: 00005-4270130723.png (1.89 MB, 1248x1608)
1.89 MB
1.89 MB PNG
>>
>>8317489
So good

>>8317516
I feel like I can hear the skin against that floatie lol


Any recommendations on what to do with pics after generation to make them look more amateur? So far I've tried upping the white balance, smoothing texture, and trying to adjust curves.

Real amateur pics outside of OF just has this certain level of jank I can't achieve due to how clean the lighting is in SD even with LoRAs.


Also coherent txt2vid soon pls.
>>
File: 00077-1164367589.jpg (1.55 MB, 2000x3200)
1.55 MB
1.55 MB JPG
having a skill issue finding the middle ground on face loras and facial expressions, it's all or nothing
>>
File: 00090-3606533991.jpg (1.74 MB, 2000x3200)
1.74 MB
1.74 MB JPG
>>
File: 00011-843346086.png (1.92 MB, 1608x1248)
1.92 MB
1.92 MB PNG
>>
File: zonkey-refinertest.jpg (754 KB, 2304x1403)
754 KB
754 KB JPG
>>8317227
zonkey+refiner. Why didn't I think of that.
See pic for a comparison. zonkey has a bit of that unnatural texture (even with DPM++ 2M SDE and CLIP skip 2 it doesn't fully disappear at 30 steps).
Running cyberrealisticPony as the refiner clears that right up. The hard lighting remains. cyberrealisticPony doesn't prompt as well as zonkey, so I'm not using that in the first place.
Running cyberrealisticXL as the refiner fishes the harsh lighting (it's a non-pony model). juggernaut and irisLux are the same, but look a bit worse.
Fucking wow. It's all the zonkey without looking zonkey...
>>
>>8317516
catbox this girl pls
>>
>>8315074
Instead of slowly getting into it, please quickly get out of it

You're dogshit
>>
>>8317737
>rrealisticPony as the refiner clears that right up. The hard lighting remai

what sampler/cfg scale are u using on zonkey?

in my testing, dpm++ 3m sde have good results, but i dont like this sampler
>>
>>8317762
Post your gens for comparison great master
They were not bad for a first try. You are just wrong
I didn't even run adetailer on the face her
>>
>>8317812
What I mentioned there, and CFG 4. Copied those from some civitai image. I'll try 3m SDE
>>
File: refiner1.jpg (219 KB, 768x1344)
219 KB
219 KB JPG
>>8315074
Image redone with better face and crPony refiner as mentioned here >>8317737
On closer inspection, a non-pony refiner can butcher the lewd details noticeably. So I'll live with the harsh lighting for now.
>>8317762
Better now?
>>
I would like to try zonkey, but I have seen that it works differently from conventional pony models, I think it does not use the Score and uses a prompt of "real life" or something like that I have understood, for the theme of shadows and lights that you recommend, what I have seen most has been "dim light, dramatic shadows", any other recommendation, is there any prompt that improves the quality of the skin?
>>
File: refiner1b.jpg (219 KB, 768x1344)
219 KB
219 KB JPG
>>8317854
I think you need both, since it's a merge. Leave out the score and you get crap. But you need some realistic elements in the prompt too.
Try it with a refiner as was just explained here, works wonders >>8317737

>>8317845
even better face (god that stupid angle)
>>
>>8317854
>>8317737

My go to has been ponyRealismv7 checkpoint, but i am reliant on loras for realism. This one plays well with everclearPNY at 1.0 weight and with ponyrealism at 0.6 weight.

https://civitai.com/models/119667/sdxl-william-eggleston-style
>>
>>8316408
https://litter.catbox.moe/bi6if3.png
https://litter.catbox.moe/ddd61s.png
https://litter.catbox.moe/v7pwsm.png
First image with refiner and fixed face:
https://litter.catbox.moe/ufkk9n.png
>>
File: 635132.jpg (1.41 MB, 2048x2048)
1.41 MB
1.41 MB JPG
>>
File: 00082-1819691666.png (515 KB, 576x768)
515 KB
515 KB PNG
>>8317216
pixiv will automatically search through your metadata for celeb names so i'd imagine civit would eventually do the same. i got hit for Mila Kunis in the early days

>>8317227
i'm not merging shit lol this is degenerate enough as is without spending even more time on it

>>8317283
yea xl needs lower CFG,idk why the slider even goes up to 30 when 99.9 percent of the time you need between 1 and 7

>>8317568
real photos have a jank to it that SD averages can't achieve. you can look at this and be like "damn i wish that were upscaled" but when you see an upscaled version, then what? is some curiosity satisfied? do you spontaneously nut? what is the point of any of this?
>>
>>8318157

It’s definitely a just a curiosity satisfied, even when using 3D modelers and things like that I always challenge myself and see just how believable I can make an image or piece of work… like if I saw it without any context that it was an AI photo that I’d believe it.


It’s just one of those things where either way you go, whether it be upscaling the photo for fidelity or purposefully adding noise and artifacts, that there’s always going to be that sort of “look” to it that is so hard to shake. It’s just a fun game and time waster to me.
>>
File: 00015-988094926.png (2.34 MB, 1352x1800)
2.34 MB
2.34 MB PNG
>>
>>8317568
Fun idea: try running the image through img2img with "amateur" in the prompt.
The clip interrogator can identify an image as amateur. That means it man also generate an image to look that way (whatever that exactly may entail).
>>
File: 00024-3507177504.png (2.71 MB, 1352x1800)
2.71 MB
2.71 MB PNG
>>
File: 00026-391153064.png (2.34 MB, 1352x1800)
2.34 MB
2.34 MB PNG
>>
>>8318027
thanks!
>>
>>8317737
Cool! What percentage are you setting the refiner at for your samples?
>>
File: 00028-2449564124.png (2.51 MB, 1352x1800)
2.51 MB
2.51 MB PNG
>>
>>8318209
It reminds me of the photoshop fake days. If you wanted to make something look really "real", you had to add fake jpeg noise to the shooped head. If the base shot has crappy artifacts, your brain expects so see those in the face as well.
Now we have new problems. I'm suspecting that some issues people are having with XL is bad prompting and AI being too sneaky. For example, Fooocus puts "noise" in the negative prompt in its style. Guess what, noise can be needed to make an image look good. And something can be mistaken for noise that is then filtered out and hurts the final image, like hair fuzz or skin pores.
And they also put "ugly" in the negative. Guess what's tagged ugly in the dataset: skin imperfections, out-of-symmetry appearances, blemishes, untidy makeup. All things that make something look real.
>>
>>8318315
0.8 like anon suggested, which is also the default value in A1111 and Fooocus and probably good for technical reasons.
>>
>>8317216
you can search on desuarchive heres a search just for emilia
https://desuarchive.org/aco/search/text/emilia/page/1
and one for Daenerys
https://desuarchive.org/aco/search/text/daenerys/page/1/

You should have plenty to choose from.
Otherwise I'll post some of her shortly.
Looking forward to your new model.
>>
File: 00033-1098247678.png (2.49 MB, 1352x1800)
2.49 MB
2.49 MB PNG
>>
File: 00034-2220672771.png (2.53 MB, 1352x1800)
2.53 MB
2.53 MB PNG
>>
>>8318353
did you inpaint the hand? it made a little hand-in-hand and arm in there
>>
>>8318372
uh bruh its fine da vaselino filter makez it gud dun fosus on detayzz
>>
>>8315575
prompts?
>>
>>8318372
oops lol no i've been fucking around with 1.5 and tile upscale

>>8318380
>nogen
>>
drawfags are mad tonight lol
>>
>>8318576
>tardgen
>>
>>8318331
Yea true i was hoping to avoid the sifting lol. Ik there were some people who made a ton of emilias and used to catbox all the time. Looking back at all those early dalle threads is depressing tho. It was fun
>>
File: 02538-2306615315.png (1.13 MB, 768x1152)
1.13 MB
1.13 MB PNG
i dont even see images anymore. just see potential
>>
File: 56456345y.png (1.62 MB, 1024x1024)
1.62 MB
1.62 MB PNG
>>8318445

it's a 3d render that I img2img with realismengineXL
>raw high-resolution extremely-detailed real natural gritty atmospheric closeup candid amateur iphone photo, terrified crying weeping sweaty dirty naked 30 year old Caucasian woman wearing torn medieval dress, natural large saggy breasts, natural nipple, auburn hair, being grabbed by several medieval warrior men, medieval warrior men tearing off woman dress, on a medieval city street, natural light, harsh afternoon sunlight, skin details, natural skin, imperfections, foggy, moody, hazy lighting, shot with iphone, natural detailed face, ugly, gritty, grimy, intense, visceral, bad photo
>>
>>8318323

That’s so funny because that reminds me exactly of what it was like lol
>>
File: aco.jpg (101 KB, 800x1411)
101 KB
101 KB JPG
the last thing you see before the probe goes in
>>
>>8318592
Source of the said salt farm?
>>
File: Roman.png (2.33 MB, 1024x1536)
2.33 MB
2.33 MB PNG
Made in SD
>>
File: 74675467.jpg (1.44 MB, 2048x2048)
1.44 MB
1.44 MB JPG
>>
File: 00078-73023448.png (1.31 MB, 1024x1024)
1.31 MB
1.31 MB PNG
>>
File: 00075-1180770219.png (1.15 MB, 1024x1024)
1.15 MB
1.15 MB PNG
>>
File: 1692135692293703.png (2.77 MB, 1344x1728)
2.77 MB
2.77 MB PNG
>>
File: 1701567956595706.png (2.83 MB, 1824x1248)
2.83 MB
2.83 MB PNG
>>8319381
>>
File: 00005-4031742856.png (1.88 MB, 1432x1432)
1.88 MB
1.88 MB PNG
>>
>>8319387
Do you have the lora link for this pose?
>>
>>8319477
More?
>>
>>8319565
that was just prompt, no lora involved
>>
File: double.png (2.16 MB, 1024x1536)
2.16 MB
2.16 MB PNG
>>
File: SPSD_Kaya (179).jpg (42 KB, 512x768)
42 KB
42 KB JPG
>>8319025
Prefer the "you found too late it´s an infiltrator".
>>
File: ok.jpg (283 KB, 1248x1824)
283 KB
283 KB JPG
sdxl too blurry
>>
>>8319917
had some good results with setting adetailer to run over the skin by setting the detection model to person and marking "use separate checkpoint" with a good SD1.5 one in there (and a prompt like highly detailed skin bla bla)
You'd think this wouldn't work since the resolution of 1.5 is lower but it actually does a good job
Makes me question XL more and more. The higher resolution is not worth much if the details get blurred out
>>
File: 00194-2635926775.jpg (175 KB, 1024x1536)
175 KB
175 KB JPG
>>
>>8320016
WOW incredible realistic JennaOrtega. Can you do ScarJor or EmiliaClarke??
>>
File: ec1.jpg (310 KB, 768x1344)
310 KB
310 KB JPG
>>8320030
NTA but I have some emilias lying around from a test
>>
File: ec2.jpg (311 KB, 768x1344)
311 KB
311 KB JPG
>>8320075
>>
File: ec3.jpg (331 KB, 768x1344)
331 KB
331 KB JPG
>>8320079
>>
File: ec4.jpg (375 KB, 768x1344)
375 KB
375 KB JPG
>>8320083
>>
File: ec5.jpg (376 KB, 768x1344)
376 KB
376 KB JPG
>>8320089
>>
File: ec6.jpg (391 KB, 768x1344)
391 KB
391 KB JPG
>>8320092
>>
File: ec7.jpg (272 KB, 768x1344)
272 KB
272 KB JPG
>>8320102
>>
File: 00062-1463702189.png (2.13 MB, 1432x1432)
2.13 MB
2.13 MB PNG
>>8319609
>>
>>8320075
>>8320079
>>8320083
>>8320089
>>8320092
>>8320102
>>8320103
SURREAL REALISTIC bravo anon, incredible work.
>>
File: ec8.jpg (367 KB, 768x1344)
367 KB
367 KB JPG
>>8320272
bonus content then
>>
File: ec9.jpg (352 KB, 768x1344)
352 KB
352 KB JPG
>>8320375
>>
>>8320375
>>8320380
Any nudes or with man?
>>
>>8317845
>>8317861
Still ass face
>>
>>8320455
Yeah maybe. But I can make hundreds more similar images without lifting a finger. And I'll keep making them and they will get better as I learn.
>>
File: b01.jpg (241 KB, 768x1360)
241 KB
241 KB JPG
>>8320412
No men. But I want to run some tests of light, poses and facial expressions and might as well do it with her at the (nudist) beach
Nevermind any garbled hands or toes since I'm only running adetailer on the face for this
>>
File: b02.jpg (324 KB, 768x1360)
324 KB
324 KB JPG
>>8320586
>>
File: b03.jpg (291 KB, 768x1360)
291 KB
291 KB JPG
>>8320588
>>
File: b04.jpg (282 KB, 768x1360)
282 KB
282 KB JPG
>>8320589
>>
File: b05.jpg (285 KB, 768x1360)
285 KB
285 KB JPG
>>8320590
>>
File: b06.jpg (271 KB, 768x1360)
271 KB
271 KB JPG
>>8320592
>>
File: b07.jpg (275 KB, 768x1360)
275 KB
275 KB JPG
>>8320606
>>
File: b08.jpg (258 KB, 768x1360)
258 KB
258 KB JPG
>>8320617
>>
File: 00022.jpg (1.75 MB, 2048x2048)
1.75 MB
1.75 MB JPG
>>
File: 00425-3433088825.jpg (210 KB, 1536x1536)
210 KB
210 KB JPG
>>
File: 00220-755566270.png (977 KB, 840x1064)
977 KB
977 KB PNG
Why is this thread on a cartoon board?
>>
>>
>>8321220
because it's adult and there is a cartoon companion thread
the celebs would fit in more on /tv/ or /hr/, but /tv/ doesn't fit the adult theme and /hr/ requires higher resolution images than the default gens provide
>>
>>8315060
QQ anyone here uses NAI and has a referral to give, please?
>>
File: 00329-755566270.png (1.03 MB, 840x1064)
1.03 MB
1.03 MB PNG
>>8321253
The cartoon/sdg thread has gens of cartoon girls.
When I think of a cartoon companion thread, I imagine an irl gens of cartoon girls
but this thread contains random girls.
Is /aco/ the best board for a thread like this?
>>
>>8321398
the cartoon thread has randos too
it's not perfect but what the hell
there's also the dall-e thread
it's much faster moving and most of it not even adult
>>
>>8318818
thx
>>
>>8316704
>>8316706
>>8316706
amazing, thank you
>>
>>8321398
you are not discouraged from posting your gens
if you can make more of these and post them, they belong here
make it your thread, don't try to change other people's posting habits with words
>>
File: 00104-1362213441.png (1.14 MB, 1024x1024)
1.14 MB
1.14 MB PNG
>>
File: 00106-660703028.png (1.13 MB, 1024x1024)
1.13 MB
1.13 MB PNG
>>
File: 00167-325762602.png (1.19 MB, 1024x1024)
1.19 MB
1.19 MB PNG
>>
>>8321398
I agree but there is no where better for it currently. DESU at this point I think we need a dedicated AI board
>>
DallE LORA uploaded. No keywords, makes good Emmas.

>https://civitai.com/models/522298?modelVersionId=580296
>>
>>8322157
Disgusting.
>>
>>8322113
>>8322157
Sweet! Thanks mate gonna grab it now
>>
>>8322157
>>
>>8322271
>>
>>8322275

>>8322164
get gud
>>
>>8322113
is the lora made to create emma? I have tried it and I like the shadows effect it adds from 0.3-0.7, but it also modifies the face a little bit making it look like emma, right?
>>
>>8322474
yes it was only trained on emma pics
>>
Woow Love that Emmas! Im looking for more! Could anyone know where i could find?
>>
File: 00172-3921071832.png (786 KB, 1248x1608)
786 KB
786 KB PNG
I'm still testing with zonkey, it's a bit more complex to use than other models, because of the hires. fix, the promts without scores, having to use the refiner to get better effects, and the adetailer, but of all I've tried is the one that gets better shadows and lighting, the other models are very good (ponyrealism, fennfoto and novareality are my favorites), but are still very "bright" and have little shadows, or are very difficult to achieve, this photo is using what I learned, with a hires.fix of 1.4 and adetailer, what do you think?
>>
File: 00177-686552945.png (759 KB, 1248x1608)
759 KB
759 KB PNG
>>8322704
For reference, this is the same prompt, only using Fennfoto, another of my favorite models, as you can see the shadows and lighting changes a lot and that's what I mean by "bright" (I didn't use adetailer this time, I just wanted to show the lighting and shadows),
>>
>>8322113
sneaky! let's hope no one snitches this time
>>
wow dalle3 is over, now zonkey is my best friend.
>>
>>8322113
>DALL-E LoRA
Just use dramatic angle and extreme angle shot in the prompt.
>>8322704
>>8322735
Yes, seeing it. The first one is a more interesting image.
>>
>>8322704
pretty gud
so the upscale factor is 1.4 and you just set Latent upscale or is there something else to recommend?
The first time I tried using upscaling it ate all my VRAM and crashed my system so I'm a bit hesitant to experiment myself
>>
File: 00001-1852839393.png (2.14 MB, 1248x1608)
2.14 MB
2.14 MB PNG
>>
>>8323109
incredible , 10 years from now you cant tell real pics from fake.
>>
>>8323115
10 years from now, we are going to have videos
>>
I had kind of a crazy idea I call it Friezafication. Orginally it was just bdsm pictures of Frieza but I thought it was hotter if this was actually Frieza's original story. He is a normal person who gets transformed into Frieza.The helmet is a brainwashing device to turn him into an insatiable cocksucker. The tail is a giant robotic dildo, his crotch is just an ambiguous mound because it's just latex or armor over a chastity cage. etc.

Somebody created Frieza and then unleashed him onto the universe. And it's no wonder he's so pissed off and also so gay.
>>
>>8323138
>>
>>8323142
eventually his skin turns white from whatever they're feeding him, and the seems on his helmet and restraints start to fade and just blend into his skin
>>
>>8322113
Thank you, you show the way working with dalle images, don't stop with emma
>>
>>8323144
until one day eventually he wakes up and he's free to leave. his captor is gone. the fucking machine he was strapped to is gone, and he needs to seek out into the universe for some cock to crave his insatiable appetite.
>>
>>8323138
>>8323142
>>8323144
>>8323151
cool stuff
meant as someone for inspiration to try make a realistic version or did you post in the wrong thread?
>>
>>8323138
>>8323142
>>8323144
>>8323151
>>>/d/
>>
File: 2024-06-18_13-04-27_2607.png (3.23 MB, 2304x1664)
3.23 MB
3.23 MB PNG
>>
>>8317605
try to make the base with a lower face weight until you get the expression
it will then have lower likeness of course
set adetailer to inpaint the face with the face lora at higher weight to get the likeness back
if the denoising strength is low enough, the facial expression will be retained
>>
>>8323022

im using the ESRGAN_4x with 0 steps,I have tried several steps, but I don't notice much difference.

i have 8gb vram and with forge i dont have any problems with vram
>>
File: 00009-2514905798.png (2.33 MB, 1432x1432)
2.33 MB
2.33 MB PNG
>>8323115
thanks, really getting the hang of zonkey, mostly by just stealing a bunch of danbooru tags...
>>
>>8323199
I just love the images of veiled Muslim women. It's so beautiful!

So, I don't know if you could give some pointers. Prompt, model, loras, etc...
>>
>>8323222
deepboru interrogate of any existing images (drawn or photos) is really helpful to learn tags too
>>
>>8322271
>>8322275
>>8322113
thank you, can you make loras with orhers celebs? zendaya, scarjo, taytay, miley, aguilera with dall-e images?
>>
>>8323223
I use autismMix to create a base image to work with, then I use it with pyracanny or image prompt on fooocus with Pony Realism then inpaint details, lora isnt needed usually
>>
>>8323215
Thanks, I tried that and 1.75 upscale to 1792x1792 works at least, with VRAM to spare
I have 12GB but the AMD driver is apparently really bad at dividing the memory between system and SD
>>
>>8323250
EMILIA CLARKE ??
>>
there's some gens in here with obvious underage girls
remember that this is a realistic thread
maybe we should link to /b/ AI DEGENerated in the op
those will fit right in there
>>
File: 2024-06-18_15-32-01_1044.png (3.23 MB, 2304x1664)
3.23 MB
3.23 MB PNG
>>
File: 00014-3141818669.png (2.13 MB, 1248x1608)
2.13 MB
2.13 MB PNG
>>
>>
>>8323223
that is the workflow I started with as well, but manual inpainting gets boring real fast
fortunately you can automate the process with adetailer
that's if you can't prompt or lora/embedding generate the required look in the first place
might be hard to combine with naked/lewdness, and loras bring their own issues
but I bet it's real easy to just prompt for a consistent hair style that has the size and color of a veil
set adetailer to detect face + dilate mask (or try a head detection model that includes hair)
inpaint prompt headscarf with appropriate denoising (can even be done with a different checkpoint, can mix SD1.5/XL/pony in one gen) and maybe headscarf lora
and presto. a perfect image every time at the push of a button
>>
File: 00021-4172522667.png (2.13 MB, 1248x1608)
2.13 MB
2.13 MB PNG
>>
File: 00047-3582061504.png (2.16 MB, 1248x1608)
2.16 MB
2.16 MB PNG
>>
File: 1712657400670347.png (2.32 MB, 1344x1728)
2.32 MB
2.32 MB PNG
>>
File: 00139-3775516251.png (1.44 MB, 960x1440)
1.44 MB
1.44 MB PNG
>>8315060
am I degenerate enough, boys?
>>
>>8323222
Can you get longer pussy hair?
>>
>>8323215
if you leave the upscaler steps at 0, it uses the same number of sampling steps as the normal gen
You can set this to a lower number to speed up the process
>>
File: 00066-1541447885.png (2.34 MB, 1432x1432)
2.34 MB
2.34 MB PNG
>>8323594
sort of.
>>
>>
>>8323721
Not bad, can you get it consistently? Or did it come with the 40 year old MILF porn star body?
>>
File: 00082-2795096552.png (2.23 MB, 1248x1608)
2.23 MB
2.23 MB PNG
>>8323813
excessive_pubic_hair gives pretty consistent results no matter how much weight. no 70s bushes for me.
>>
>>8323843
>that
>excessive
LMFAO let me guess, tags from a pedo site?
>>
>>8323222
>>8323435
nice one, guy. you found the sweep spot for genning. it looks good face looks good.
>>
>>8323853
danbooru tags, so I guess.
>>
>>8323858
You don't want that, I just looked at the list of tags and it has practically nothing relevant.
>>
File: 00075-4271357876.png (2.2 MB, 1248x1608)
2.2 MB
2.2 MB PNG
i tell I lie, a weight of 4.2 in A1111 gets you a proper bush.
>>
>>8323889
This looks like when people who never saw a pussy started making Skyrim skins.
>>
File: 00097-906561254.png (2.28 MB, 1248x1608)
2.28 MB
2.28 MB PNG
>>8323889
oops, pre-inpainting.
>>8323886
ah, I'm not really familiar with danbooru. I'll look for something better.
>>
>>8323211
good idea, that’s almost what I’ve been doing except I didnt change the adetailer prompt to a higher strength lora. I’m going to be building a new rig soon, so I should also be able to upscale/inpaint much more quickly.
>>
File: 00106-1108851338.jpg (590 KB, 3000x4000)
590 KB
590 KB JPG
>>
File: 00117-942854616.jpg (653 KB, 3000x4000)
653 KB
653 KB JPG
>>
File: 00169-1570292267.jpg (582 KB, 3000x4000)
582 KB
582 KB JPG
>>
File: 00014-3548740720.jpg (396 KB, 3000x4000)
396 KB
396 KB JPG
>>
File: 00015-2756420606.jpg (419 KB, 3000x4000)
419 KB
419 KB JPG
>>
>>8323957
that's the way to go
>>8323989
saved
>>8323993
pretty good
>>
File: 00022-2596382525.jpg (388 KB, 3000x4000)
388 KB
388 KB JPG
>>
File: 00031-3216414279.jpg (419 KB, 3000x4000)
419 KB
419 KB JPG
>>
File: 00049-1059727495.jpg (473 KB, 3000x4000)
473 KB
473 KB JPG
>>
File: 00056-3674749093.jpg (476 KB, 3000x4000)
476 KB
476 KB JPG
>>
File: 00100-1350500261.jpg (438 KB, 3000x4000)
438 KB
438 KB JPG
>>
File: swiss half elf.png (1.04 MB, 768x1024)
1.04 MB
1.04 MB PNG
I got deported here from /g/ despite censoring.
Here are some catboxes I posted at request, seems better than spamming the thread.
https://files.catbox.moe/nzoch8.png
https://files.catbox.moe/1n7hnu.png
https://files.catbox.moe/g4pve0.png
https://files.catbox.moe/4sad5m.png
https://files.catbox.moe/08n5qt.png
https://files.catbox.moe/zd4ozs.png
https://files.catbox.moe/i44onf.png
https://files.catbox.moe/vpf5xn.png
https://files.catbox.moe/h6zf94.png
https://files.catbox.moe/e0pble.png
https://files.catbox.moe/i23ss0.png
https://files.catbox.moe/rjj2i5.png
https://files.catbox.moe/1semc0.png
Someone please tell me why realesrgan_x4plus upscale doesn't actually work and eats the original image but produces nothing ever when I enable upscaling in metastable, even though it is their recommended upscale module.
More details on the problem here.
>>>/g/101040207
>>
>>8323271
>>8323468

Ok nice, thank you a lot ! =)
>>
>>8323897
How long does inpainting a small part like that take for you compared to generating an entire image? I have a shitty PC so it would take me less time and effort if I just shooped a good looking pussy from a porn pic.
>>
File: 00064.png (1020 KB, 768x1024)
1020 KB
1020 KB PNG
>>8324265
>>
File: tableau eghehghe.jpg (1.08 MB, 1500x996)
1.08 MB
1.08 MB JPG
Redoing paintings with img2img can produce good results
>>
File: tableau 02 gegee.jpg (452 KB, 1072x774)
452 KB
452 KB JPG
Another one.
>>
File: yshtola ai 007.jpg (224 KB, 1152x1440)
224 KB
224 KB JPG
>>
File: 00091.png (821 KB, 744x1024)
821 KB
821 KB PNG
>>8324337
>>
>>8324344
>>8324357
Do you use an autotagger on the source image for the img2img or something? Whenever I img2img it generates a totally new image with some similarities but different positions and stuff
>>
File: 00093.png (832 KB, 744x1024)
832 KB
832 KB PNG
>>8324516
>>
File: 00095.png (822 KB, 744x1024)
822 KB
822 KB PNG
>>8324529
>>
>>8324519
NTA but lower your denoising strength
>>
>>8324519

So I use img2img mode.

- Model: RealVis XL 3.0.

And in my prompt, I put: Realistic analog amateur photography posted to facebook, Fujifilm XT3, Canon R5. Realistic, photorealistic.
I often add a few details of the image in the prompt to make it easier (Girls, happy, naked, etc.).

I also put a negative prompt: EasyNegative, anime, cartoon, 2d, 3d cgi, bad anatomy, clothes, bikini, dress, polydactily

I use the sampler: DPM ++ 2m karras.

Yes, I use an adetailer: face_yolov8s.pt
>>
>>8323155
>>8323168
i specifically put it in his thread on this board.theyre made in stable diffusion so this is the right thread and they arent anime style, theyre western style so its aco not d. not sure what is difficult to understand about that.
>>
>>8324600
but we are the realistic SD thread
there is a separate cartoon SD thread
what is difficult to understand about that
>>
>>8324621
/d/ is for eastern art and full of gay pedos
/e/ is for eastern art and full of pedos
/h/ is for eastern art and full of pedos
/p/ is for actual photos
/hr/ is probably the actual right place for a realistic/photographic thread
/aco/ is for western art related to /co/ and /tg/ and is the right place for threads of stuff like >>8324265
there is no faggot/futa western art board which is a FUCKING mistake
photofags/realismfags should go to /hr/ and start up a colony there
the deranged repulsive fag you replied to should unironically start a general on /lgbt/, I'm sure those screechy twitter pedos will cry about AI but will be outweighed by the other coomer faggots
I wish faggots just wouldn't fucking come to /aco/, it was such a good board back when especially the futafags were over in /aco/
>>
>>8324650
this thread here is not what's shitting up your board though
that's the dalle thread
>>
>>8324684
I'm not complaining about this thread though, or that one, I'm just talking about what each board is actually for.
What's shitting up /aco/ is trapfags.
>>
File: 00132-3199397597.jpg (1.62 MB, 1920x3072)
1.62 MB
1.62 MB JPG
doing a lot of face experiments - I suppose now I know a lot of ways NOT to make a lightbulb
>>
File: 00088-2906522048.jpg (1.45 MB, 2080x3040)
1.45 MB
1.45 MB JPG
and the damn freckle clumps keep popping up
>>
File: 00222-3186648700.jpg (1.55 MB, 2080x3040)
1.55 MB
1.55 MB JPG
>>
File: 00046-3984857855.jpg (1.79 MB, 2240x3200)
1.79 MB
1.79 MB JPG
>>
File: 00054-465550916.jpg (1.65 MB, 1664x2432)
1.65 MB
1.65 MB JPG
speaking of this being a cartoon board, I messed around with 2.5d for like five minutes... drawn styles are definitely easy mode
>>
File: 00002-2488772729.jpg (128 KB, 768x1280)
128 KB
128 KB JPG
>>
>>8323119
rather 1 to 2 years
>>
File: 00285-231875811.png (1.65 MB, 960x1440)
1.65 MB
1.65 MB PNG
>>8315060
damn, this shit looks like renders
>>
File: 00055-2463189725.jpg (139 KB, 768x1280)
139 KB
139 KB JPG
>>
File: 00155-2929132002.png (2.57 MB, 1152x1536)
2.57 MB
2.57 MB PNG
>>
>>
>>8325311
If you add a head that's this clear, you gotta get rid of the zonkey grit everywhere
Try cyberrealisticPony as a refiner at 0.8
>>8325248
In this one it's barely noticeable
>>
>>8325672
FUCKING HELL MAN , anon this is scary how realistic it is. bravo. Can you do it with Emilia Clarke or ScarJo?
>>
>>8325723
>Can you do it with Emilia Clarke or ScarJo?

Not at the moment; what you're seeing is the result of a LORA that was hammered together in my trusty forge over the course of several millenia.
>>
>>
>>8324784
if you are having weird troubles, try simplifying your prompt
the AI works by association. seemingly unrelated terms can affect the outcome in unexpected ways
for example, prompting brown eyes changes the whole face and body since those are associated with different ethnicity
there is probably something hostile to ginger/freckle types in your prompt (positive or negative) without you realizing it
I wish there were a tool to find these hidden interactions
>>
>>8325978
make sure to kick out "ugly" from any of your negatives
it's the opposite of beautiful, which is associated with clean skin, airbrushed fashion photography glamor crap
>>
>>8325738
>course of several millenia
Spend couple of millenia on EmiliaC and ScarJo because anon that head of JennaOrtega is incredible, fucking spooky realistic
>>
File: 4155104403.png (1016 KB, 768x1280)
1016 KB
1016 KB PNG
>>
>>8326097
Very nice pic.
>>
File: 00021-3948194309.png (2.18 MB, 1248x1608)
2.18 MB
2.18 MB PNG
>>8324322
>inpainting
the area you're inpainting doesn't matter, as it is done at the resolution settings you put in, so a smaller selection will be processed at a higher resolution and get more detail. generally takes me about 30 seconds each time with a 6700xt using SDXL, but varies with model, steps, resolution. Unless you're a master at photoshop, you're not going to be quicker. I used a 1050ti with SD1.5, generation times were about a minute.
>>
>>8326186
6700xt here too
not to mention that if your denoising strength and padding is set right, it automatically matches the light and camera angle to whatever is needed
and if you are not happy with the result, let it make a batch of 10 and grab a coffee
it's literally automated photoshopping
>>
File: 00031-3789111371.png (2.23 MB, 1248x1608)
2.23 MB
2.23 MB PNG
>>8326199
>6700xt here too
pain
>>
File: 1695646521337558.png (2.66 MB, 1440x1632)
2.66 MB
2.66 MB PNG
anons, would you consider this gen blurry?
>>
>>8326343
pretty high detail for a latex doll
>>
File: 1710557389621982.png (2.87 MB, 1440x1632)
2.87 MB
2.87 MB PNG
>>8326343
>>8326358
oof
how about this one (ignore the vacuum face)
>>
File: 1718807181879282img2img.jpg (358 KB, 1128x1280)
358 KB
358 KB JPG
>>8326343
cute face, but the skin...
>>
File: 1718809308614418img2img.jpg (355 KB, 1128x1280)
355 KB
355 KB JPG
>>8326441
>>
File: 1701586453208331.png (2.45 MB, 1440x1632)
2.45 MB
2.45 MB PNG
>>8326460
>>8326472
I see your point anon
how about this one
>>
>>8326484
this one is oddly perfect, just run adetailer on the face and/or eyes
>>
File: 1713473440581191.png (2.56 MB, 1440x1632)
2.56 MB
2.56 MB PNG
>>8326484
>>8326491
thanks for the feedback anon
turns out the workflow I use for 2D was backfiring big time for 3D
>>
>>8326512
now her face looks a bit rubbery again
but there is actual skin detail
these are all not bad looking by the way, just different looks/styles
>>
File: 00319-1659181603.jpg (148 KB, 1152x1920)
148 KB
148 KB JPG
>>8325700
>>
File: 00335-1808549191.jpg (176 KB, 1152x1920)
176 KB
176 KB JPG
>>
File: 2024-06-13 20_33_19001531.jpg (1.13 MB, 1725x2153)
1.13 MB
1.13 MB JPG
the thread has been so milquetoast mid that I had to step in.
>>
File: 2024-06-13 16_04_47001497.jpg (1.07 MB, 1606x2003)
1.07 MB
1.07 MB JPG
>>
>>
>>
File: 2024-06-14 19_02_12001579.jpg (1.38 MB, 1871x2157)
1.38 MB
1.38 MB JPG
>>
>>
File: 2024-06-13 16_23_36001518.jpg (1.13 MB, 1711x2151)
1.13 MB
1.13 MB JPG
>>
File: 2024-06-13 16_31_16001522.jpg (1.07 MB, 1724x2148)
1.07 MB
1.07 MB JPG
>>
File: 2024-06-13 20_34_54001533.jpg (1.3 MB, 1725x2150)
1.3 MB
1.3 MB JPG
>>
File: 2024-06-14 19_01_03001576.jpg (1.1 MB, 1547x1891)
1.1 MB
1.1 MB JPG
>>
>>8326720
>>8326748
cool. looks almost too clean heh
but head and body are a perfect match

>>8326774
you sat on these for almost a week, what a welcome gap filler
>>8326785
hot as hell, is there a keyword for choking/throat grabbing?
>>
File: 00475.png (1.85 MB, 1344x1352)
1.85 MB
1.85 MB PNG
AOC back in the community college/bartender days
>>
File: 00474.png (1.98 MB, 1272x1400)
1.98 MB
1.98 MB PNG
>>
File: 00476.png (1.91 MB, 1272x1400)
1.91 MB
1.91 MB PNG
>>
File: 00473.png (1.94 MB, 1312x1312)
1.94 MB
1.94 MB PNG
>>
File: 00478.png (2.42 MB, 1528x1528)
2.42 MB
2.42 MB PNG
>>
File: 00480.png (2.43 MB, 1528x1528)
2.43 MB
2.43 MB PNG
>>
File: 00479.png (1.74 MB, 1040x1384)
1.74 MB
1.74 MB PNG
>>
>>8318592
U mad tho
>>
>>8320272
What are you smoking those are all dogshit
>>
>>8323591
Not even remotely, we need to go deeper
>>
>>8327085
POST BETTER THEN????
>>
>>8326960
>>8326963
>>8326964
>>8326967
>>8326973
>>8326977


Dollar store knockoff AOC
>>
ignore dismissive nogen comments, we all know who's behind those
>>
>>8327103
I'm waiting for a guide on how to set things up
>>
>>8327117
pathetic
>>
>>8327114
Biden stfu and go back to being charged for felony gun possession
>>
every AI thread has a resident shitter
in here, the cartoon thread and even in dalle (take yer meds...)
I wonder if it's the same guy actually
>>
File: 2024-06-14 18_14_55001561.jpg (1.2 MB, 1724x2149)
1.2 MB
1.2 MB JPG
>>8326929
I was banned :(
I use "choking" + subject + weight, (too much of a weight and you get anomalies, invisible hands etc.)
>>
>>8326781
good fucking lord
>>
>>8326790
holy
>>
File: agent4.jpg (1024 KB, 1228x1228)
1024 KB
1024 KB JPG
>>
File: agent5.jpg (1.11 MB, 1228x1228)
1.11 MB
1.11 MB JPG
>>
File: greetings4.jpg (1.08 MB, 1152x1440)
1.08 MB
1.08 MB JPG
>>
File: on_tap.jpg (1.05 MB, 1152x1440)
1.05 MB
1.05 MB JPG
>>
File: hangin_out.jpg (911 KB, 1400x1400)
911 KB
911 KB JPG
>>
>>8326781
You got a discord? Take commissions?
>>
>>
>>
>>8327119
Retard
>>
>>8327192
Finally a good post
>>
>>
>>
>>
>>8327153
thanks
overdoing it with the weights is a universal truth
you just can't force things too hard, if the AI cannot place something and you force it with high weights it starts bending reality to fit the prompt
>>
File: ch1.jpg (257 KB, 768x1344)
257 KB
257 KB JPG
So she is in demand
Should have said something sooner
>>
File: ch2.jpg (290 KB, 768x1344)
290 KB
290 KB JPG
Remember to wear protection
>>
File: ch3.jpg (238 KB, 768x1344)
238 KB
238 KB JPG
>>
File: 00071-1719529913.png (939 KB, 768x1280)
939 KB
939 KB PNG
From my testing so far, I can't get Zonkey to look nearly as good as PonyRealism. (Picrel is Pony, not Zonkey)
Zonkey seems so be too noisy and the faces are always fucked for me. I did use a refiner, but it didn't seem to help THAT much and it took too long to gen things. Skill issue?
>>
File: 00045-1953193996.png (1.07 MB, 768x1280)
1.07 MB
1.07 MB PNG
>>8327582
>>
File: 00061-988756027.png (1.03 MB, 768x1280)
1.03 MB
1.03 MB PNG
>>8327591
>>
>>8327582
Did you lower the CFG to 4? At 7 the grime is so crazy the refiner can't fix. Also try CLIP skip 2 and then try another sampler
>>
File: 00026-1175641957.png (930 KB, 768x1280)
930 KB
930 KB PNG
>>8327599
I tried several different CFG values based on sample images I saw, and I usually do CLIP skip 2
>>
>>8327612
And samplers? DPM++ 2 or 3 M SDE is usually recommended, some are terrible
>>
File: ch4.jpg (293 KB, 768x1344)
293 KB
293 KB JPG
adetailer pussy detection works like a charm
>>
>>8327635
missed some pubes damn
the inpaint mask is not the only thing thats gonna get dilated
>>
>>8323889
You can use "happy trail" for more hair towards the belly.
>>
File: ch5.jpg (325 KB, 768x1344)
325 KB
325 KB JPG
should be no more stray pubes and switched to better pussy prompt and larger tits
all automated by adetailer, no manual steps
>>
File: 21609-4167656338.png (2.17 MB, 1344x1728)
2.17 MB
2.17 MB PNG
>>
>>8327943
That's a waste of perfectly fine milk.
>>
>>
>>8328003
>>
>>
>>
I have about 1k pictures of a similar model.
>>
>>8322113
can anyone re-upload it please to catbox or something?! thanks. really wanted this one :/
>>
File: 21638-3494959303.png (2.19 MB, 1344x1728)
2.19 MB
2.19 MB PNG
>>
>>8327943
>>8328453
a literal titcow
>>
File: 00036-1335854770.png (1.33 MB, 1024x1024)
1.33 MB
1.33 MB PNG
>>
File: 00024-1573601821.png (1.63 MB, 1024x1024)
1.63 MB
1.63 MB PNG
>>
File: 00073-1932292434.png (1.17 MB, 1024x1024)
1.17 MB
1.17 MB PNG
>>
File: 00011-104548593.png (933 KB, 1024x1024)
933 KB
933 KB PNG
>>
File: 21663-3231042520.png (2.29 MB, 1344x1728)
2.29 MB
2.29 MB PNG
>>
>>8327943
>>8328453
>>8328547
fresh milk at the table, love it
>>8328512
>>8328517
maximum realism
>>8328519
is this inspired from some really really old bondage scan? the face and size of the tits are really specific, and the pose matches

we page 9 now, guess I'll bake a new thread
>>
New thread

>>8328842
>>8328842
>>8328842
>>
bump



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.