/vt/ - Virtual Youtubers

Request anchor

>>39628772
gimme a really nice fauna with the nakedribbon embedding faunanon i know you will deliver

Merry Christmas and enjoy the rest of the year, /vtai/.

>>39628834
whats wrong with OP? : (

>>39626762
>how much emphasis is usually too much?
I'm hesitant to go above 1.3 unless I'm really fighting with the AI to get it to do what I want. Generally I do first attempts with nothing more than 1.1, which is what you get if you don't put a number, and then turn parts of it up more if I need to.

Just because something is a central feature of the art you're aiming for doesn't necessarily mean you need to emphasize it a ton. If you're trying for something very specific/rare, or have a super long prompt (especially when parts of it contradict each other) you'll probably be throwing emphasis everywhere but otherwise it's usually best to not overuse it in my experience

https://litter.catbox.moe/t39jjs.png

>>39628901
she has a "finger" coming out of her legs. This one is really good tho

>>39628901
there's a giant fleshlight wrapped up with her

>>39628657
fuck. I spent the last twenty minutes finding one of these that would work for "/vtai/ - VTuber AI-generated Art - all I want for Christmas is OkayuouououuuOUOUUuuuu edition" and flipped all of okayuposter's mariah okayus. WHY DIDN'T YOU CROP THEM. GRID HATE. the hat fit perfectly too. my disappointment is immeasurable. you've ruined christmas

>>39629063
>>39628657

>>39628657
It looks like Fauna has a giant futa cock in that pic and I think that's cool

>>39628931
I find 1.4 is a hard limit I dont breach.

>>39629050
>>39629057
oh shit

>>39629088
>>39629063
The early bird gets the worm okayuposter

>>39629132
I'm not Okayuposter. if the OP image was good I'd have not said anything.

>>39628772
Extra squishy Kronii but with Tifa clothes

https://litter.catbox.moe/fr3gdg.png

>>39628657
Why does Fauna have a giant deformed futa cock?

>>39629206
Squish comes later, I wanted to do a normal one first.
It kinda does a mix of Kronii's paizuri gear and Tifa's wifebeater.

>>39629419
>>39629132
how's that worm taste?
lmao, naked ribbon made her mariah outfit even closer

>>39629694
You probably want her closer to obese than this I guess.

>>39629866
forgot to say the magic words for nightmare eating, "spaghetti, noodles, ramen", kek. and to turn the snark up to 11 with "mr. early bird". it's just japes, merry christmas faunanon
>>39629947
nta but that's pretty nice and squishy. make one fit like Tifa and pencil leg Kronii? just one. don't go Schwarzenegger on me pls, just toned

I love it when the AI fights itself to do conflicting shit and just makes sex outfits instead

>>39630180
the embedding was mainly trained with the navel showing so it will try to force it

plus tits that aren't glued together in something tight. loose clothing, breasts apart, plunging neckline might be here to dethrone skintight/skindentation for me.

>>39626836
thank you!
>just a naked ribbon embed
unfathomably based
>nene hypernetwork
damn, I do want to play with this stuff but any real GPU time I have right now is rented.

>>39630248
??? I'm not using any embeddings at the moment. shitposting from NAI directly is still the most convenient.

>>39630248
just realized you thought I was talking about the naked ribbon embed. nah I meant I just threw the NAI tag in here for something that was already asking for very specific clothing.

>>39630403
yeah nai alone gets confused with that tag and if you add bound, bdsm or similars it swaps it for rope, if you dont mind paying you can always use runpod even tho its more expensive but at least you get the power

>>39630467
I'll have to check that out. I have a colab instance that I can run with either the regular nVIDIA T4 or an A100/V100, but just waiting the 5-10 minutes for webui to launch costs me as long as running the former for an hour when I decide to use the latter, kek

at this rate I'm going to be proompting christmas maririns until easter. I waited way too long to start playing with Christmas-y tags.

>>39629132
>The early bird gets the worm okayuposter
Yeah I'm just getting up.
>>39629063
WHY DIDN'T YOU CROP THEM. GRID HATE
I mainly just go with grids until I can find one that I like and then I'll refine it and post it seperately. Gomen.

>>39629694
Oh wow, shredded is nice too!
Thanks!

>>39629947
Actually not even, that's pree much the level of squish I had in mind, maybe a teensy bit squisher but not much more, and that's already very nice! Thanks for that!

>>39630532
i've never felt the need of using anything more than a 3090 and if you wanna be cheap you can also use spot at the risk of your work being interrupted

I just got another long message from a Japanese person on pixiv, but this one does not seem to be please concern for the children. Quite the opposite actually.
お忙しい中、すみません。用件だけ項目に分けて伝えます。
1.pixivにあるイラストをアイコン、ヘッダーや、加工(少しの加工、トリミング程度)、○○○可愛い!とかネタ画像をTwitterつぶやきで一緒にイラストをアップロードなど(その他のつぶやき)して使いたいのですがいいでしょうか?
2.また、アイコン、ヘッダー、加工以外にアップロード時に関しては基本的に出典表記をして使おうと思っているのですが、名前(Twitter ID表記)、出典表記が出来ない垢で使うことがあります。その際は例として、「その画像はどちら様の画像でしょうか?」という風に聞かれた場合は、DM(ダイレクトメッセージ)などで作者様のお名前、Twitter ID表記、出典表記をお伝えして対応しようと思っています。商業目的はありません。もし分からないことがあればなんでも聞いてください。
Deepl: I am sorry to bother you. I'll tell you what I'm looking for by dividing it into the following items.
1. I would like to use illustrations from pixiv as icons, headers, processing (just a little processing and cropping), uploading illustrations of XXX cute! or uploading illustrations together with Twitter tweets (or other tweets), is this OK?
2. In addition to icons, headers, and processing, I would like to use the source of the image when uploading, but there are times when I use the image with a name (Twitter ID) and a source that I am not allowed to use. In such cases, I will ask, for example, "Which image is that? I will respond to such a question by informing the author's name, Twitter ID, and the source of the image via DM (direct message) or other means. There is no commercial purpose. If you have any questions, please feel free to ask.


Am I stupid or is there no downside to this since I don't upload my stuff to twitter anyways?

>>39628772
Nina Claus wishing (You) a Haggy Holidays.

>>39629990
I always have some difficulty getting moderately toned outputs, the AI can do slight chub, proper muscled or 80s action heroes well.
A tasteful hint of abs? Good luck.
Maybe I'm missing some better prompts, but toned seems very similar to "muscular female".

>>39630800
you need to use athletic and toned and not use muscular at all or she will become hulk

>>39630800
I just do "toned stomach" and it seemed to work alright.

>>39630640
He's just asking permission to do stuff anyone in the western world would do without asking, because who the fuck cares

>>39630852
Japanese people are too polite.

>>39628772
requesting an extra thick chuuba sneaking downstairs to eat all the christmas cookies!

>>39630640
Yeah sounds like he just wants to use your stuff as an avi
I just get Japs complaining about bush kek

>>39630827
>>39630823
I think these kinda work. I'll play with them more in the future.

>>39631019
do they not like a bit of jungle down there :/?

>>39628772
can anyone that has access to that robot model try the nakedribbon embedding?

>>39631019
kek I gotta post some bush on mine eventually.
Are you the mr. upscaler kronster tatsunoko high-roller man by any chance?

why do I still have more fun coomer-grugging tags and excludes with a locked seed on NAI than I do with webui?

>>39630467
>runpod
>Community Cloud Pricing
>$2.040/hr A100
Jesas. That's works out to twice as much as it costs to use an A100 on Colab - It uses ~10 compute units/hr, and 100 of those were $9.99
>1x Tesla V100 16 GB 0.280
>1x V100 SXM2 16 GB 0.290
but, Google might stick you with a V100 when you request a "premium GPU" instance. and I've never used one of those, so I don't know if the rate goes down by 10x or not. The compute units work out to ~$0.20/hr when you're using a regular T4 GPU on Colab, so using an RTX A4000/A4500/A5000 on runpod might be worth it. Have you used it before? I'm looking at their Docs and I would much rather not learn more python, nor any Docker at all, kek. They have "spot instance" prices that are supposedly cheaper, but they also charge $0.2/GB/mo for idle storage
I might play with it next month, but I have persistent storage in Drive from a family member's Google One, so it is very convenient. A bit slow as shit to launch, but once it's up, whether I turn the Colab off or Google does, I lose nothing, the images are already safe.
>>39630571
I already had to crop it to flip it vertically, kek
>>39630640
Reply with,
>すごい!ぼくはバカ外人です、日本語本当に難しい
j/k. sounds like he wants to use them as reaction images or emotes or something?
pro-tip: sometimes DeepL shits the bed harder than it normally does with long/compound sentences. separating them out or running them individually can sometimes give better context. it can also make it terribly worse. here's another attempt of 2.
2.
Also, I'd like to use icons, headers, and

I am going to use it with the source notation basically when uploading it except for processing.

However, there are times when I use them in a manner that does not allow me to indicate the name (Twitter ID) or the source of the photo.
In that case, if someone asks, for example, "Whose image is that? I will respond to such a question by providing the author's name, Twitter ID, and the source of the image via DM (direct message) or other means. There is no commercial purpose. If you have any questions, please feel free to ask.
>>39630800
what >>39630823 said. "muscular" = Hulk. I just do "fit/athletic/sporty one, not all three, groin tendon" and it always gives me what I want.
>>39631064
SEX. I MISS THE GUY THAT WANTED NAVEL PLAY. HE MISSED SO MANY HAPPY ACCIDENTS

there any good references for artists to switch up the style of images?

>>39628772
Chubby Sana showing off her two sweaty armpits, with steam, intense blush and a smile

Good shit or secret sauce?
https://litter.catbox.moe/pprhae.png

>>39631300
I like Shit better but they're both good.

>>39631202
>why do I still have more fun coomer-grugging tags and excludes with a locked seed on NAI than I do with webui?
Probably because it's faster. It's hard to beat iterating on a seed dozens of times in a span of a minute.

>>39631202
i use it daily and the storage per moth isnt an issue since you can just download whats important to your pc and for fast uploads you can upload your models, etc to different drives and use gdown. For example today i set up a new pod in like 10 minutes and i had to download like 7 models, embeddings, hypernetworks, etc. I havent tried google collab but i head they made it more expensive or smth

>>39631344
Yeah Okayu comes really well on good shit.

How tf am I supposed to fix backwards hands on the webui?

>>39631438
You draw her hand in an image editing program. That's what I did for the Suisei.
Then you run it through img2img to fix it up.

>>39631465
Gotcha, I'm trying to figure out how best to use inpainting in the webui. I used MSpaint to fix some Towa hands last night...
What image editing program do you use?

>>39631497
Paint.net is a happy medium since it's fast and powerful.
Stuff like Gimp is overkill and takes too long to open.

>>39631525
Gotcha, thanks!
I've also noticed that hands seem to get worse on higher denoising strengths.

>>39631211
I tried to do a steam censors but the AI was having none of that.
So I went with pasties, but she still her clit hanging out sometimes.
This was probably the most tame output I got...
I'll just trash the rest. Too lewd.

>>39631300
I only played with secret sauce very very briefly, but I found that it makes hands even worse than usual. and I didn't proompt NSFW because elGoog would've had a copy forever.
didn't faunon mix some gape model and other stuff in there? i.e. it would be my own choice instead of NAI-full, if Santa shows up drunk with a GPU next week, but I'd use good shit for anything NAI-curated could've done.
>>39631345
it really is. I fucked off as soon as I saw your response and made 70 of these trying to make the next image you'll see from me in a minute or so

>>39631675
God that's so fucking hot

>>39631345
sex

>>39631357
can grug figure out how to use it easily if grug absolutely abhors docker? I see they have a Jupyter...thing, which means I could use the Colab notebook I already wrote and connect it to runpod as a runtime instance?
>>39631525
it depends on what you're trying to do? GIMP might be overkill, yeah, but I've actually considering using Photoshop for a few of mine for it's AI magic wand/selection and healing brushes. they're fucking ridiculous.
also skill issue. it launches in...hang on I'll time it. 17 seconds on the overgrown potato I use for CPU image gen. off SATA300 SSDs. in a cave. with a box of scraps. I just keep it running there anyways.
>>39631638
honestly half the time I use MSPaint tho. nothing against paint.net, but I already knew how to use PS, and then "learned" GIMP and got my brush presets and everything for using it on other machines before I got the uber hacked Photoshop to work reliably everywhere. and promptly dumped the technically legitimate but "would've instantly deactivated itself the second it was able to phone home" install I still had of edu Creative Cloud

>>39631753
I haven't actually opened Gimp in years so maybe it's faster now.
Paint.net does have the magic selection though.

>>39631068
Tragically
>>39631199
Not me, sorry

>>39631678
Agreed, I think with good shit for most general use but I definetly want to give it a whirl on some NSFW stuff later.
>it really is. I fucked off as soon as I saw your response and made 70 of these trying to make the next image you'll see from me in a minute or so
Yea in the span of you doing that I've interated on one Okayu seed like, 14 times. I wonder if the beeg brain (and big $$$) play is to use the NAI website to quickly refine seeds and then do the gigachad upscaling.
>>39631753
I've never used paint.net before so if I can't figure that out at least I can fall back on good ol mspaint.

>>39631675
hmmm now thats an embedding i could try to work on cause i never get the steam censor right

>>39631849
>Not me, sorry
Nothing to be sorry about, it's hard to be keep track of who's who around here. I'm the dum dum.

>>39631809
GIMP does too. Photoshop's new one is on another fucking level. The magic scissors that snap to edges work better there than they do anywhere else. It's not fair. More than 1/2 of the time GIMP just spent launching was it loading extensions/Python. If you won't be using any of them, just move them out of the extensions directory. I ought to figure out where they auto-load from and just back them up, I never use them.
>>39631876
I think it's supposed to be easier to learn than GIMP or PS. This was the only thing I ever actually bothered inpainting outside of NAI's interface, because I wanted to select by color and then make two separate sets of outlines.

>>39631876
>>39631997
which made these. I ought to try latent upscale for this on good shit, but I think I'm going to try and "plan ahead" for paid compute time and just have a list of tasks to do.

>>39632091
these looks amazing anon love the colors

Daily wife 2

https://files.catbox.moe/hu81f6.png

If i go schizo enough i'll try to make an embedding just so it makes decent looking knifes cause the ai shits itself

>>39632122
thanks. posted them about a month ago, I think the request or maybe even just banter question was "can it make senketsu bibi colored?" and I made it so.
>el tema del día en pixiv es "negro"
I know what I'm doing, kek
>>39632169
sex. props for not just making a half-rate 2B like most of the shitters I saw when I checked earlier
also don't let Ayame near them once you're done.

>>39631997
>>39632091
That's honestly some of the coolest stuff I've seen someone do with the AI.
>I ought to try latent upscale for this on good shit,
What's latent upscale?

>>39628657
cock

has anyone had any luck with a realistic mix that can still handle replicating the girls?

>>39632196
i feel like you would go ham if i succesfully made a lam, mikapikazo model that worked and i hadnt checked what people had been doing but honestly it doesnt surprise me kek

>>39632198
https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/5521
TL;DR it lets you manually get something close to "hi-res fix" but on img2img upscales instead of being limited to using it initially with txt2img, which opens the door for things like using "hi-res fix lite" after you inpaint, or after switching models, or changing the prompt a bit, etc.
someone PR'd a commit that gives you a new option on the img2img tab - which normally does any upscaling in "image" space. Instead it tokenizes the image again and does the upscaling in black magic AI CLIP "latent space".
https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/5840 you can also use it in x/y plots now too
https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/5976 - I've been using this on my Colab btw, I thought it was always there but I guess I was mistaken. it's quite nice, but the previews are "your VAE is off"
>>39632306
>lam, mikapikazo
nani sore?

>>39631751
Have another one.

>>39629947
Now I'm wondering if a kick upskirt is possible, this Kroniifa is doing things to me...

>>39632375
just two artists that use saturated colors really well

>>39632281
you'd have to ask the wawa poster that was fucking with the various instagram models and stuff. I've done a lot of...let's call it the opposite. kek
oh shit, there's some 2B in this folder. kek. guess I'm going to become a doubleplus shitter today too
>>39632398
if they shade "flat" then I would indeed go ham. I spend most of my proompting time trying to get nice output with flat shading

>>39632558

>>39632621
hell yeah I love that shit. I can make colors pop and the shading flat pretty consistently now. kaguya luna poster gave me the gift of "vibrant colors" kek. or drab with extra flat shading and "weirdly flat reflections on skin", 'megami magazine' on either alternating steps or only on the upscale, kek, just add water. it doesn't seem to work as reliably on the various mixes, depending on how much NAI is left in there
on the other hand, consistently getting thick outlines out of this thing is notoriously difficult, and megami gives pencil thin outlines so it's out. "emphasis lines" can do it but it's a nasty prompt fucker most of the time, makes you turn scale down

>>39632375
I see, thanks! How do you add these to the webui?
t.smoothbrain

>>39632853
They were pull requests, and they're all closed, so they're "built into it" already. Shut it down, "git pull" again, and start her back up, you should see them show up.
t. evensmootherbrain most of the time

>>39632381
Kinda. Things get anatomically messy.

It's funny how much better my ass stuff does on pixiv than my boob stuff. I guess it makes sense since I'm more of an assman myself so I guess I know better what to pose and what to look for. For boob outputs I feel like an alien trying to blend in with humans.

>>39632974
actually that's really good, whatever that is supposed to be in her mouth is bothering me but other than that this turned out really well

I feel like i would get more followers if i posted the nenes that i post on /h/ but i do it for fun and i feel like getting sfw good looking images is actually harder

I'm debating making a new pixiv to post stuff or just using my existing one.

>>39633097
Tried to improve her face a bit since you liked it.

>>39632884
Thanks a bunch for explaining it to me in a way my grug brain can understand. This will make upscaling my old shit much easier!

>>39633257
What would be the differences in what you would post on the two pixivs?

>>39633262
would it be too much trouble to do white undies instead of black?
sort of like classic Tifa
I think she looks good in the outfit desu, the massive thighs and underboob work to her advantage

>>39633348
Well I don't have anything on my current, but some people know the account and it might be awkward posting stuff on it.

oh euler...

>>39633492
Gotcha. If it would make things awkward then you might as well make a new one dedicated to what you would post from here.

12 anlas were left in here. exactly enough for one upscale at 0.62 noise. it is indeed Christmas. four perfect fingers.

>>39633079
kek, nobody likes the kind of shit I have absolutely the most fun doing: 混ぜるな危険
they're all my worst performers. and when I forget it, somebody adds the tag within the first hour.
>>39633332
No problem!
>This will make upscaling my old shit much easier!
It will certainly make them look better, if you use the latent upscale option it's not that far off from using hi-res fix. The PR had some comparison images I think.
I just wish the batch img2img option had a way to read the prompt/negative prompt information from each image, keep the sampler/steps/resolution/cfg the same as what you set them or even lock them individually, like keep the sampler I choose, but maybe read the cfg from the input image, and then process the batch. Also, IDK if I'm retarded or what, but batch img2img writes NO metadata.
>>39633348
>>39633492
NTA but my sisters are aware of mine and I already fret about posting stuff that'd make me enable the "suggestive" flag. If I post R18 stuff I'll have to dump it on exhentai or make a second one too...
>>39633507
I had very bad luck with denoising above 0.6 with latent upscale on. Or with trying to exceed 1152/768 in either direction with it on.

>>39628772
Not a request, but a question. Assuming VRAM isn't at all an issue, is there a technical "upper ceiling" to how much you can upscale/hi-res fix an image before it goes to shit? I was trying to make WUXGA backgrounds yesterday and they were ALL garbage. I actually wanted to do WQXGA 2560x1600 but the slider in the UI stops you at 2048. It still had VRAM to spare too, but the images were beyond shit at 1920x1200 already.

>>39633398
Done. Also fixed up her odd side boob part, that I hadn't even noticed initially.
And also her suspender that was super thin in one spot.

>>39633616
check the training data of your model. Most models aren't conditioned to generate at a base resolution that high which is why you get garbage
the common models will have bad image composition starting at around 1024 in either dimension
I use the upscale feature for what you've described

>>39633903
I wasn't trying to generate at that res, the first pass was never bigger than 768x768. Usually 512x768 or 768x512. Just any kind of upscaling couldn't really go beyond 1536 pixels wide or tall.

>>39633867
Nice! I really like Kronii in this outfit, think it works really well, thanks for doing this!
I noticed the side boob but sort of thought it was supposed to be like a bit of chub seeping out of the tank top since it's so tight

Merry Christmas, anons.

>>39633616
>Question
So I dont think there is a problem with highresfix itself but rather with landscape images in general.
NAI was trained with aspect ratio bucketing which makes it theoretically possible to generate with higher than 512x512 resolution and not have too much trouble.
See https://blog.novelai.net/novelai-improvements-on-stable-diffusion-e10d38db82ac
However going higher than 512x768 (portrait) for firstpass seems to produce model errors because its trying its best to fill in the empty space.
I havn't found a firstpass that works well for landscape images when trying to generate characters lying down on the side that doesnt deform their stomach badly. Even going down to 640x384 doesnt produce great images either. I think its down to training data and the model trying to make characters stand up preferably.

I think its probably better to try to generate at a smaller widescreen resolution and then latent upscale it atleast 2x for the added details and after that pull the image to extras and give it a final pass there to your desired widescreen resolution for your wallpaper needs.

There is an option to save the image before applying highresfix in oiptions that could be useful to further understand your problems.

>>39634186
The initial lower resolution gens were all fine, and I could upscale without latent upscale as high as 1536x1024 with no issues on most of them. I wasn't trying to make them lie down either. I was trying to do it in one go though, I hadn't thought of e.g. doing two back to back latent upscales, or maybe trying latent upscale on something I'd already upscaled without it.
>>39634058
Channel toryoku shittene! Merry Christmas!

>>39634186
nvm, I never tried it without the schizo negative, just checked the original image to see what I was doing

>>39633961
I just tried and got it to do the highres upscale to the 2048 limit fine (24GB VRAM)
somewhat blurry due to the huge jump in resolution though. I still think SD isn't adding any value at these resolutions do it's better to do a small bump and then another upscale separately

>>39632380
Oh my, thank you!

>>39634284
Pulled yours through extras with crop to fit

what prompts would I need to get metal cat ears? I've tried fake animal ears, headgear, ears made of metal, grey ears, and a number of other stuff but only got regular ears or hair tufts that are vaguely ear shaped.

>>39634414
wtf, what/how did you do? still looks a bit waifu2x-y but it's vastly better than what I was getting trying to use the AI to img2img upscale
a̶̪͓͒ḧ̵̡̛̪͎͉̻͎̥͓̖͔͔̞͉́̏͂͛̔̈́̒͑̀͘͘͝ͅō̴̡͙͖̘̖̗͇̖̠̞̳̩̟̣̱͕͑͠y̴͔̥̮͐͛

>>39634414
That was with R-ESRGAN 4x+ Anime6B
You can see it sharpens the blurred stuff a little too well
This one is with lanczos

>>39633616
>kek, nobody likes the kind of shit I have absolutely the most fun doing: 混ぜるな危険
Tragic. People on pixiv seem to have little care for AI stuff that is not lewd.
>they're all my worst performers. and when I forget it, somebody adds the tag within the first hour.
At least your people know how to find you!
>I just wish the batch img2img option had a way to read the prompt/negative prompt information from each image, keep the sampler/steps/resolution/cfg the same as what you set them or even lock them individually, like keep the sampler I choose, but maybe read the cfg from the input image, and then process the batch. Also, IDK if I'm retarded or what, but batch img2img writes NO metadata.
Yeah I haven't tried batch img2img yet but if it doesn't write any metadata at all that's bogus. Img2img really works best with iterating on a single output anyways.
>spoiler
Yeah my sister knows about mine too because I can't keep my mouth shut. She doesn't know who I am, just that I have a pixiv with AI stuff.

I had very bad luck with denoising above 0.6 with latent upscale on. Or with trying to exceed 1152/768 in either direction with it on.
Yea I mainly use the DPM samplers now. I just didn't realize the AI put me back on Euler after I updated lol

>>39632169
You could also just inpaint it enough times until it looks half decent, but it takes some editing of the prompts and stuff.

https://litter.catbox.moe/xjt05u.png

>>39634490
Black magic : )

>>39634490
your denoise is too high

>>39634597
Also not sure if any of these matters much but here are mine

>>39634439
beats me. I just manged to kinda sorta tricked it into doing it with cat ear headphones
>>39634539
it is also useful if you're trying to play with latent mirror but don't want it still applying during hi-res fix. doing a batch img2img after using latent mirror on smaller images is the easiest way to do that. I tend to use DPM++ 2S a Karras or DPM++ 2M - haven't tried Karras for the latter yet. you ever feel crippled by choice? first thing I did was fucking hide half of them.
>>39634597
>>39634667
>Lanczos
kek, I use that shit to downscale video when I'm not using Spline64. who knew it was that easy?! thank you!!! do I go by what's selected in the dialog bc the Lanczos under the preview is erroneous, or did you change it after?

>>39634657
How high do you go? With hi-res fix at lower resolutions I can get away with using ~0.7, but usually 0.6. What you replied to was 0.69 nice.. Picrel was 0.5. Still nightmarefuel.

>>39634853
Damn bro, I literally use denoise strength of 0.05 to 0.1

>>39634872
Wait what? I use strength from like .4-.7.

>>39628772
Suisei or Towa, wearing nothing but a pair of high heels and a sweater, sitting with her legs crossed next to a yule log fire.

>>39634960
From what I've noticed, higher denoising strength usually changes the image for me, however slightly, and adds in too much of a smoothing effect to the lines, which imo makes it look worst.

>>39634556
i mean i could draw it myself if i really wanted to put in the effort.

>>39634872
>0.05
damson. I kind of let the upscale add detail. in fact I tend to hope it does. I'll try going way lower next time then, 0.2 or something.

Has anyone tried making something with an AMD instead of RTX? Everything I read about AI says it doesn't work with AMD for whatever reason

>>39634853
the point of using SD in highres/img2img is to get the model to add finer detail to the image
since it is bad at adding details without destroying the composition at higher resolutions, you should choose a very low denoise value, less than 0.2
if you think the image is already detailed enough at ~HD resolution then you should do the final upscale with an upscaler instead

>>39634810
>cat ear headphones
yeah that seems to work better. cheers.

>>39634853
>>39634872
>>39634960
>0.1
thinking emoji

>>39634995
When I leave my strength that low I get stuff like this. Do you keep it between .05-.1 just for upscaling?
>>39635052
>Has anyone tried making something with an AMD instead of RTX? Everything I read about AI says it doesn't work with AMD for whatever reason
lol
yea I tried. That's why I got an RTX.

>>39635129
Yeah that low only for upscaling is what I do.

>>39635087
ive been using 0.65 for so long and it usually gives me better results

>>39635087
>>39635129
he's trying to go to 2k in one dimension
if you have the denoising that high at that resolution it's going to put in random nonsensical crap (similar to if you try to just generate at 2k base)

>>39635087
you gotta take into account it follows the original image.

I use 0.6 personally for highresfix

>>39635087
I don't think we are talking about the same thing. I was referring to 0.05 to 0.2 denoising strength when specifically upscaling. Not when generating image with highres fix. My bad.

But this is an upscaled image of IRyS to 4x original resolution at 0.1 denoising strength.

https://litter.catbox.moe/4hd7p3.png

I'm gonna eat now. Then I'm gonna come back and pick my best Mariah Catto's and mess with them.

>>39635055
I'd hardly call 512x768 HD though. also I realize I'm not talking about wallpapers anymore.
the added detail worked out pretty nicely: https://litter.catbox.moe/811vu5.png
except for the one where I got the ghost of AI-hands-mas past to show up, kek: https://litter.catbox.moe/las5sp.png
>>39635087
I'm not sure how or why 0.7 is sticking out as somehow much better than values both higher AND lower than it, wtf was going on here?
>>39635182
oh. her arm was melting in the source. denoise isn't just upscaling here then, kek, it's still drawing anew!
>>39635251
it looks very nice***
*** zoomed to fit at 1280 height
it looks kind of bad at 100% zoom it looks terrible... - I guess it's useful for super-sampling? go that big and then go back down until it's not nearly as blurry/smooth anymore.
BUT I am very thankful for this knowledge. now I know I can still hi-res fix up to "the limit" which seems to be 1.5-2x the "ideal dimensions" so 1536x1024/1024x1536 and then use low denoise strength to get all the way up to e.g. 1920x1200 or something. I imagine it'd still look less melty/smooth/waifu2x than going straight to the extras tab.
>>39635303
Bon Appetit

thank you all for explaining to grug. and sharing your knowledge.

>>39635251
So img2img Just resize(latent upscale) or SD Upscale under scripts?

Gotta stitch my SD upscale attempts together cus I cant run x/y and sd upscale at the same time

I miss her so much bros...

>>39635502
SD upscale script is what I'm using. The specific upscaler I'm using is called lollypop

Ah yes, Aigiya Kuuna, my favorite vtuber.

>why is it taking so long
>opens resource manager
>oh

>>39635542
Share your settings?

>>39635630
These settings?

>>39635680
Yeah.
Except I think we're trying to compare latent upscale and sd upscale together

>>39635372 (my dumb ass)
plus there's one thing I still haven't been able to wrap my head around: "Noise"
I know the combination of the two values pictured = ENSD 31337 for txt2img using ancestral samplers, but NovelAI actually lets you adjust them independently when you img2img. and they make results change a lot.
>>39635542
Hang on, we're talking about different things entirely now...I thought you were saying to use 0.2 denoise for a purely img2img upscale, not SD upscale script?

>>39635953
>>39635542
ah. I was just about to ask. I've never used that before.
>>39635843
and yes, I was originally asking about latent upscale or hi-res fix, either or going to complete shit once you approach 2x of 512/768 on either dimension, because regular img2img kept it together AT 2x most of the time, even with 0.5-0.7 denoise.

>>39628772
Maid Coco mad at (You) for forcing her to wear such a short outfit with white panties if possible

>>39635087 (txt2img highresfix)
>>39635502 (img2img latent upscale)
>>39635843 (img2img (latent upscale + sd upscale)

Personally the ones with highresfix and sdupscale and latent upscale doesnt fuck her legs at all
Picrel is with just resize and no sd upscale with 0.6 denoise
Apparently this chucgs VRAM kek

>>39636000
Remember to select an upscaler (?)

>>39635372
.7 feels like the sweet spot for me as well. With one exception...
(((HANDS))))
Sometimes at .7 it will add fugly hands where there were none. And if there are hands it makes them fuglier.

>>39628772
Nenechi in her shoulderless Santa outfit

>>39636117

>>39635537
The day I get a real GPU is the day I never stop prompting those Jupiterian tits

>>39628772
Christmas started off okay but it turned out the one thing I was looking forward to all day is something not happening by forces out of my own.

Feeling crummy anons, can someone cheer me up with cozy laplus by the fireplace? Maybe eating cookies or something

>>39628772
Mio losing her sanity as she hears "the turkey will be done in just a few more minutes" for the 8th time this afternoon, the chef's voice the only sound drowning out the constant minecraft sound effects from both of her cousin's tablets at full volume. If we don't eat soon I'm going to intentionally start an argument so I can leave. I hate family holidays.

>>39636133
Yeah I know, I'd have to download the lollypop model they're using to play with it. Or try the included ones out.
>>39636195
yeah, that was the ghost of handsmas past I got too.
>>39636117
small world, I already had a coco maid that was very angry

>>39636133
Just resize + sd upscale @ 0.6

Its still very potent way of upscaling if setup right. Just uses a little more VRAM than i'd like but 4x is a lot too.

>>39628772
Kiara wearing Fubuki's outfit.

>>39636117
>>39636390
kek, she's angry at the absurdity of her outfit

whats the rule of thumb for using the hi res fix? like how much smaller should the first pass be of the final image?

>>39636514
https://media.tenor.com/zx7jlZbNxpwAAAAC/goten-sneer.gif

>>39636556
I always use 512x768/768x512 for the first pass, and I haven't had much luck with letting hi-res fix go beyond 1152x768/768x1152 - 1.5x. I've also never really had any luck generating first passes at anything much bigger than the aforementioned values, but that probably has more to do with how NAI was trained.

>>39636390
Yeah if i'd take away anything from this is that using SD upscale is pretty good. The settings arnt that difficult. You can see the legs shitting themselves whenever I dont use it.

>>39634961
AI kinda gave up on the sweater in this picture, but I think it's the best one I managed to generate.

>>39636651
Alternatively, one where the AI gave up on the fireplace.

>>39636673
fancy
>>39636117
there we go. believe it or not the smaller I made her tits, the less I had to {{{pantyshot, skirt lift}}} to actually get her to lift the damn thing

>>39636195
Funny you say that because 0.2 - 0.5 here have decent hand, but 0.6-0.7 is fucked, then fine again at 0.8.
https://litter.catbox.moe/pdlhoc.png

>>39636556
This >>39636611
But I stick with 1024x1536 and it works fine for me.

>>39636335

>>39636258
>>39636390
>>39636532
>>39636765
very cute! thank you! I like this art style

>>39636850
Yea this checks out. 0.6-0.7 do something really fucky with the hands compared to the others.

>>39636884
>The tie Lou1!
What did she mean by this?

>>39636890
not sure how me and the first anon arrived at the same thing at the same time, I just remembered this old coco maid when I saw your request and made her pantsu white, lol. if you follow her back to warosu/alice I posted a few others in the same style too

>>39634961
Here you go

>>39637245

>>39636850
0.8 is easily the best here. 0.2-0.5 look too "waifu2x". what's this - latent upscale, SD upscale, both?

>>39637379
Just highres. fix when initially generating image at 768 x 1024 or 768 x 1280. I forgot which resolution.

>>39637370

>>39628772
Sana in a tight leotard.

>>39637459
>>39636651
>no lineart
>is very noticeably lineart
What did prompter mean by this?

>>39628772
A 4-panel comic strip of Fubuki shapeshifting into a hamburger.

>>39637379
>>39637417
I think you gotta take into account what the original image looks like before highresfix or upscaling.
If your denoise value approaches 0 it will respect your original image the most and produce an almost 1-1 replica.
But it doesnt add in any extra detail so it will look very bland and simple.
Higher denoise around the 0.5-0.7 sweetspot, (experiment to find what works best for you) still respects the original composition enoughish while still being liberal enough to add in a lot of new finer detail with the image.
Higher denoise >0.8+ will produce something completely different.
Whats the point of upscaling one image if you get something completely different as a result? At that point you start losing control over the model.

>>39637602
"No Lineart" has the word "lineart" in it, which will make the AI generate an image with lineart.

>>39637602
It works better when you're not crowding it with other clashing vectors.

>>39628772
I have lactation fetish, do me a favor

>>39637710
I think you're missing the point for mine. I generate image with HighRes Fix on. So the generated image to me, IS, the "original image" that you're speaking about.

Usually when generating image that's around 768 x 1024 or bigger, without highres fix on, depends on what model, embedding, and hypernetwork you have on, it can generate some whacky image where there are multiple heads, legs, arms, elongated bodies, etc.

So by turning high res fix on while generating image at that resolution or higher, I get an overall more complete and "normal" looking composition, therefore, that's an "original image" for me.

>>39628657
Thefuck is that giant hog between her legs? God I fucking hate AI art.

>>39637978
Yeah thats the whole point of highres fix.
Generate at a lower resolution and apply some AI latent space magic to upscale it and keep it sameish.

>>39637558

>>39637468

>>39628772
Are Nijisanji request allowed?

>>39638525
No. We are Tribalfags.
Seriously, you can ask but only some like Ange or Sociere are doable well.

>>39628772
Scarlre wearing open toe heels (or normal ones) with pantyhose and a long sleeveless sweater while she has arms up. If it doesn't work with Scarle, do it with Elira, please.

>>39635537
I can feel my pelvis crushing by just looking at this.

>>39636314
Here u go anon, I got her drinking some coffee to keep herself warm while she waited for u :)

>>39628772
Milky IRyS

>>39637662
bonus transformation included

>>39637459
cute oyabun!
>>39639132
uooooooooh
>>39637468
>>39638495
cute twaps!
>>39637978
NTA, but
>I generate image with HighRes Fix on. So the generated image to me, IS, the "original image" that you're speaking about.
it isn't. it makes the one image, and then upscales it. but yeah, when I go any bigger than 768x768 first pass with hi res fix or in general without it I get multiple subjects, multiple asses, multiple racks, etc.
>>39638525
ofc. whether or not it can be done depends on who
>>39639043
NTA but I must know how you did this, lmfao

>>39639517
uoh, you say?

>>39638927
>>39637945
Very pretty kitty.

>>39638927
{bigger}

>>39639686
too-woah

>>39638897
Does this work?

>>39640160
Works perfectly well

Man I really want to post an Okayu set for Christmas but I dunno what to do. Okayu Mariah Carey is probably too off model for pixiv but I wasn't feeling the wrapping paper Okayu either.

>>39640160
Ma'am your areola are showing

m̵̨̳̮͎̩̪̱̺̼̥̭͇̎͐̈͗̈́͛͌̏̈́͒͒̓̕͝o̵̮̪̙̹̞̬̥̝̱̔͗̈́͘ͅg̶̜̗̐̐̈́͂̍̽͝ù̴̗͙̯̫̓͌̄̈́̏̓͠͝ ̸̨̨̧̛̙̣̝̬͖̫̙͂́̂̇͛̏͐̆̎̏̕͠͝͠ṃ̴̇̈́̉͑̓̆͊̍̓̓̚̕͝ơ̷͍̝̬͖̙͓͓̮̑̽͊̕͠ͅǵ̶̺̺͆̋͑͜͝ū̷̢̡̨̼̙̠͖ ̵̧̯̣̩͙͈̹̰̭͓̹̰͒́̇͗̔͊̒̑̐ǫ̴̝̜̤̦͎̦͔̠̫̫̾̈̊̓̆́͒̈k̶̡̛̫͍̣͕̣̫̥̟͚̿̀͋͊̋̈́̈́̉̈̈̚͜͜ͅa̴̼͓̼̼̟͉̻̬̟͇̺̤̘̐͑̕͜y̷̏̔̾͋̏̚͜ȗ̴̘̪̖̠͙͕̼̠͈͖̘̃́́̀̔͜!̷̢̢̘͚̞͈̱̟͇́̌̏̄͑̿͊̕

be not afraid

>>39638833
I like Wiwa so I'll try that in a bit.

>>39638871
Thank you anon, this brightens up my day. Laplus cute and comfy! The jacket looks nice too.

>>39640564
It's like when you're watching an old tape on an old CRT that's on it's last legs. I love it.

>>39638525
I’ve been able to grug a few like ririmu, shiina, besides those the other anon mentioned. Honestly just name a girl and I’m sure someone will try

>>39640373
Half and half set maybe?

In another universe, Kizuna Ai is alone on Christmas, touching herself while thinking of (you), only to despair after her temporary high fades and the loneliness sets in once more.

I am finally free of the nightmare. Request dump coming now.

>>39641214
I think what I've decied is that Mariah Catto will be something fun I did with you guys and the problem with with the wrapping paper set was I didn't go far enough.

>>39589625
Here you go.

>>39617800
Here you go.

>>39623553
My beautiful sweater wife.

>>39624070
Here you go. Migos in Paris.

>>39627035
Kimoi.

>>39627177
Here you go.

>>39630917
Caught her red handed.

>>39636117
Here you go.

>>39638833
Not really what you asked for but I kinda like this output enough to highres fix it.

>>39636211
She seems to want something from you.

Ina is too damn strong, enough emphasis on "tentacle rape" to actually get tentacles also gives pointy ears and squishy hair

>>39637558
Here you go.

>>39638897
>>39637766
And, finally, a two for one deal for both of you.
https://files.catbox.moe/md51c7.png
Merry Christmas, /vtai/.

>>39639517
just "fubuki, hamburger, 4koma comic" and spam generate

>>39643006
Merry Christmas, I hope you dream of your oshi

>>39643006
Literally Santa big good Mio.
Unfathomably based.

>>39628772
POV getting a nursing handjob

>>39628772
Botan wearing only a santa hat.

>>39644335
Seconding this.

>>39644711
R-18 or nah?

>>39644850
>wearing only a hat
NTA but I'm pretty sure that means she's supposed to be naked.

>>39644911
I mean like naked but covering her breasts so it's safe for a blue board or full on nude is different.

>>39644850
Whichever you feel like doing, I'm not too picky.

>>39642850
Gorgeous, thank you anon

>>39645047
Aight. I'll get to work on it soon(tm) if nobody else does it before then.
Just gotta do some laundry first.

>>39628772
Twintail FBK

>>39636335
Don't be like that anon, try incest

>>39639132
That's pretty good
>>39636884
I like this one the most

>>39645411
I forgot, glasses too

>>39644335
>>39644816
Here's a couple random Marine's that may or may not be what you're looking for.
https://files.catbox.moe/cdoa5h.png
https://files.catbox.moe/pyd51z.png
https://files.catbox.moe/4c1tyh.png

>>39645747
NTA and i genuinely mean no offense when i say this, but those cocks look more at home in a dark fantasy game's 'Gnarled Forest' area

>>39645747
I'm honestly not holding out hope since there's not a lot of POV breastfeeding art to feed into the AI anyway

>>39645411
The AI is really averse to giving her twintails for some reason.

>>39645627
Glad you like my contributions.

>>39645952
>>39646004
That's just, like, your opinion, man.

>>39645747
The second one would be okay if it wasn't for the face. Thanks for trying though.

>>39642802
I'm sorry nenechi I'm not.. strong enough

>>39638833

>>39628772
chubby chubas

>>39642228
are you using an embed here? I don't see it in the mega link for suisei.

>>39648318
Nah, thats just the NAI website. I don't use embeds for anybody in my posts.

>>39648318
>are you using an embed here?
The list of Hololive talents that need an embed is vastly smaller than the ones that don't. The ultra popular ones like Suisei really don't need one at all.

>>39648249
Ah finally. Something I can dredge out of my txt2img pile

>>39648440
Limbs sugoi. Very nice chub though.

>>39648500
I would fix the limbs but I'm already fixing the limbs on something more /d/egenerate.

>>39638833

>>39648440
Found another image.

>>39648736
Holy fuck that thigh meat.

Huh guess she shows up better than I expected for no embedding. Getting official outfits is too hard but dressing her up isn't too bad.
Guess I'll training something else.

>>39588886
where you at? I'm digging for more space wherever I can find it, and I found these dicking around in a folder from 2016, kek. any of these useful to you?
>>39641255
fucking kek
>>39642157
>unfulfilled request anon delivers
>>39642228
>>39642295
>>39642354
>>39642409
>>39642462
>>39642520
what a fucking beast
>>39643283
sugoi
>>39645747
>cute kizuna ai
>blacked.com dicks
never change vtai
>>39648318
do they grow you lot in a factory? are you using waifu diffusion? I don't get it...
>>39649009
most holos show up just fine without an embedding...

>>39649009
>Huh guess she shows up better than I expected for no embedding.
It's Suisei... You can almost grug her outfit on the NAI website.

>>39649381
not if he's after ye olde SOVLsei the one she drew herself - you'd need a SOVL artstyle embed or something just turn the vae off, kek
>>39645411
>>39646099
damn, you weren't joking. time to fuck with excludes and make it better

No stop mellowing out. Be a bitch again for christ sake

>>39649337
Beyond all odds, I actually managed to get the password. Thank you for your offer though.
And everybody should have their requests filled on Christmas.

>>39649675
part of me was hoping it was you. I'm glad
>>39649660
kek, you lost son?

>>everybody that was talking about upscaling
>let me see how this shitty CPU instance fares
>it's dogshit slow
>it actually looks fucking great
>and it's only halfway done
kek
ignore the tensor whatever error, it's from the first aborted run, CPU instance very very very unhappy if you don't use a resolution that is the same exact ratio as the source. 0.2 denoising btw.
https://litter.catbox.moe/uiva9z.png

>>39649869
>ETA 04:13:43
Damn...

>>39650042
the half baked one at 5 steps that I "Copy Image"d off the preview as fucking base64, lmfao is already my background. next time I try this I'm going to crank it a little, maybe 0.4. also DPM++ 2S a Karras is doooooooooooogshit slow on an already dogshit slow CPU, DPM++ 2M would have already finished. whatever was making my images melt on a GPU doesn't seem to happen when this pair of extra stanky fucking Xeons is doing it, lmao plus my heat is out. it's quite nice sitting next to it!!!
>>39645411
>>39646099
cute

>>39646099
and since you put her in glasses, I realized I've never tried something as a prompt. I am very pleased with the result, and I can practically fucking hear her

>>39650522
and with multiple angles, multiple views in the excludes, I CAN ACTUALLY HEAR HER AND NOW YOU CAN TOO

>>39650571
>it actually gave her proper twintails with this shit in here
I love AI

>>39650668
>>39650571
>>39650522
>>39650215
are you using an embed here? I don't see it in the mega link for fubuki

>>39650710
it worked with the other guy because he was a new IP

>>39628772
prompt your oshi with the megane pasta. no cheating and using an expression/emotion, all of this is coming from what she said, kek

>>39650904
Kek.

>>39648440
>>39648736
WHO TOLD THIS FAT HAG SHE WAS ALLOWED TO SEDUCE INNOCENT YOUNG MEN

>>39651185
If I put her name last instead it goes full schizo.

>>39628657
Which nakedribbon embed did you use?
I used 10000 twice and the result was bad

>>39651236
I didn't say you had to use the whole thing, kek, I removed parts of it so I could fit the rest of what I wanted in there with 250 tokens:
masterpiece, best quality, genshin impact, shirakami fubuki, hololive, Glasses are really versatile. First, you can have glasses-wearing girls take them off and suddenly become beautiful, or have girls wearing glasses flashing those cute grins, or have girls stealing the protagonist's glasses and putting them on like, "Haha, got your glasses!" That's just way too cute! I really like when their glasses have that suspicious looking gleam, and it's amazing how it can look really cool or just be a joke. I really like how it can fulfill all those abstract needs. Being able to switch up the styles and colors of glasses based on your mood is a lot of fun too! It's actually so much fun! You have those half rim glasses, or the thick frame glasses, everything! It's like you're enjoying all these kinds of glasses at a buffet. fox ears, very long hair, twintails, floating hair, white hair, aqua eyes, detached sleeves, side-tie peek, sideless outfit, single earring, hair ribbons, 1girl, solo

I spent way too long on this butt. Time for santa hat Botan!
https://files.catbox.moe/tmscg5.png

>>39651236
Mio's cleavage is so strong it broke the letterbox.

>>39651501
The entire thing wouldn't fit so I cut the sentence with boys and the sentence with Marine and Luna. Left me with just enough tokens for Ookami Mio, hololive,

>>39651696
Holy sex. Thanks for the catbox because I would have definitely asked.

>>39651733
>Mio's cleavage is so strong it broke the letterbox.
lmfao. sounds like we picked the same parts to cut too.

>>39651784
No problem!
>>39651696
And now that I post this I notice that her buttcheeks are different sizes...

How it was called the thing you need to run on safetensors before using them and hopefully not get your ass hacked?

>>39628772
Role exchange between Lui-nee and Mori (Lui as the grim reaper's apprentice, Mori as the HoloX executive)

Can someone do some chubby vtubers? Not super fat but still with stomach rolls. Any vtuber would be hot but especially Pomu Rainpuff or Kiara, or both of them

>>39652299
Oh hey. Something else I can dredge from my txt2img folder. Something akin to this, anon?

>>39651896
you don't need to run anything on safetensors
the tool for the other checkpoint formats was pickle scanner

>>39652461
Thank you!

>>39644711
Ey boss you still here?
https://files.catbox.moe/u4txf7.png

>>39652427
My god your Kiaras are always god tier, thanks for the squish

>>39652940
NTA but thank you for the meal
any chance you could do other chuubas in this same setting?

>>39653952
Naked with a santa hat? Yea. Do you have any one in mind before I default to Okayu?
https://files.catbox.moe/4xhuus.png

>>39654569
I don't see a lot of vshojo chuubas compared to holos, but as long as they're cute I don't care
go wild, anon, I appreciate anything you give

uhh
https://files.catbox.moe/rw02wf.png

>>39654802
HAIRIER

>>39654788
I don't have any of the vshojo embeds. Who's your favorite one of them?

In the meantime I'll try for a good Okayu.

>>39628772
Abominations from the past 2 or so days that you saved

>>39654802
SEX

>>39642625
>that face
>those thighs
>that upskirt
chef's kiss, there's something about Coco in cute outfits that just does it for me

>>39652427
This is pretty hot, the stomachs just seem too big and almost hard. Is there a way to make them just look like a soft squish? Chubby kiara is amazing either way

>>39654802
dangerously high test add some steam too

>>39655032
Hypernet accidentally set to 1 gave me a biblically accurate Fauna

>>39655032
https://files.catbox.moe/qmtf1y.png
https://files.catbox.moe/yybm2j.png
https://files.catbox.moe/tkacpy.png
https://files.catbox.moe/8mbrss.png

>>39654989
with nothing but a santa hat? definitely silvervale or mousey
but I will gladly accept okayu

Maybe it's because I've had this thing running for like 9 hours but I notice that the webui is slower now. Any way to speed it back up or should I give her a break?

>>39655695
Are you on win10 or higher? If so then Task Manager > Details > If sihost.exe is at the top with an absurd number, end task. It will restart automatically and then you should be fine.

>>39656167
Sorry, sort by memory (active) and see if sihost is at the top.

>>39656167
I am on win10. At the very top is python.exe with 9,387,068k of memory. I take it I should end that?

>>39655032
Oh hey, another thing I can dredge up from my txt2img folder

>>39656331
Oh yeah, close your cmd prompt for webui and python should disappear from your task manager, then you can just restart webui.

>>39656458
wawa if she were a centaur...and melting

>>39655032
I think I already posted most of them. Have fleshblob coco. and some very and intricately ahoy
https://litter.catbox.moe/jtris2.png
https://litter.catbox.moe/2x5e10.png
https://litter.catbox.moe/7op98j.png

>>39655032
I don't remember what happened here.

>>39655032
kek

>>39655482
Rapunzel pubes Okayu, my beloved.

>>39655532
Alright I just both embeds. I'll give them a whirl in a bit.

>>39657905
Rapunzel, Rapaunzel, won't you let down your lilac pubes?

>>39657541
Neo-Coco, boss of the extremely niche Hololive Christmas hack of FF5

>activate generate forever
>forget to unlock teh seed
pain peko

>>39658960
KEK. how big is the folder? are they really all the same image? was it at least an ancestral sampler so there might be a strand of hair that moved?

>>39659009
It wasn't too bad because I just went downstairs to refill my water. But yes, they're all the same image. It wasn't even a good seed!
It was DPPM++ 2M

New thread doko?

>>39659272
MARINEPOSTER, NOW'S YOUR CHANCE!

>>39659365
ogey. baking all I want for christmas is okayouoyuoyuyuuy since the catalog still has hats

>>39659416
Too slow, I beat you to the punch,
>>39659484
>>39659484
>>39659484

>>39659511
beat you to the anchor, kek

fucking hell, it's perfect in the thread and shite in the catalog

>>39659666
I'm glad we ran through this thread in time for this christmas miracle.

>>39659666
>>39659888
look what just came out of the oven after these back to back trips!