Thread 8944432

Previously, on /asdg/ >>8920752
The cartoony side: >>>/aco/csdg

No DALL-E, Meta AI or hailuo AI generated stuff.
No futa.
>>>/aco/futasdg
Report trolls.

>Guides
More modern options:
https://github.com/comfyanonymous/ComfyUI
https://github.com/lllyasviel/stable-diffusion-webui-forge
Too old, might hold some value: 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
>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

4chanX Catbox/NAI prompt userscript: https://rentry.org/hdgcb
4chan-x userscript from /h/ that lets you read generation metadata from images in-browser, and upload to catbox with one click when you want to share metadata.

>Wan Video 2.1
Setup Guide: https://rentry.org/wan21kjguide
Triton Guide: https://github.com/woct0rdho/triton-windows/blob/readme/README.md
Easy Sage Attention Install: https://github.com/woct0rdho/SageAttention/releases

>Related
>>>/b/degen
>>>/h/hdg
>>>/e/edg
>>>/d/ddg
>>>/g/sdg
>>>/trash/sdg
>>>/trash/slop
>>>/vg/aids
>>>/vt/vtai

My prompting is probably basic. So when a second character besides the 1girl is present, somehow the 1boy follows a bit of the open mouth or closed eyes along with her.

>>8944516
In general prompts are applied to every character in the scene, as long as there's no conflicting tag at which point you usually get one of each. Unless the innate strengths of those tags are drastically different. So you may be able to counter "open mouth" with "closed mouth" but you can't tell in advance who gets which. Anime solves this with "faceless male" which is always applied to the guy, but that doesn't work well in realism.

You can also use adetailer or inpainting to change the faces at a later point. Or use regional prompter to give each character a separate prompt that won't leak over to the other one.

>>8944516
Co-sign on >>8944526 regional prompter recommendation.

I only recently got a new computer capable of handling this stuff and its awesome to know it comes straight from the machine. Getting into animating them now.

>>8944569
May your GPU be plentiful of VRAM.

hitchhiker traveling light

welcome, enjoy your stay with us

thoroughly clean your slut after use for best results

uh ... sure, i guess we can share the tub

your grandma had a wild streak

of course, daddy

enjoy delicious kraken nuggets

magazine but you've had a stroke

Anyone have any recs for some good vintate (30s, 40s, and 50s) style checkpoints or LORA?

Was running some IP-adapter experiments and ended up with this glowing office suit.

>>8944569
Nice. I got a fairly decent rig for the purposes of genning but I keep falling in to gooning to stuff of 4Chan and testing the policy filter on Gemini. I can't focus long enough to bother getting start with sd

Gentlemen... BEHOLD! My 1girl slop

>>8945229
it is impressive that you decided, and managed, to upscale it that big. what did you use?

>>8945390
Forge for base gen, and then Krita to inpaint. It's just that big haha

>>8945431
but are you using the default hires fix, or an upscaler? what settings?

>>8945439
Yeah. Hires fix.
>1400x1600
>Hires fix of 1.15 with nearest upscaler
>Then send to img2img to upscael further
>Then send to extras to upscale one last time with Nearest + ESRGAN at 2.4x
Then into Krita to inpaint for details :)

>>8945451
thank you

>>8944745
> This issue: guest editor Hugo Ball.

What do you guys use to avoid the bias towards very bright images in the porn trained models?

>>8945897
You can try either some lighting LoRAs or add tokens like dark, ambient, indoors, vintage,

>>8945897
You can also use the Vectorscope plugin to decrease brightness and increase contrast.

>>8946208
Shit, this is great. Got anymore cyberpunk gens? I'll take anything but I'm fond of the mad moxes, super tatted up prostitutes with goth/alt aesthetic and piercings.

>>8945897
It's easier to talk about specific checkpoints and metadata rather than "the porn image trained models". I would download it and test if my suggestion works before putting it out here.

For a general approach you can look into Vectorscope or Diffusion-CG (different tools, same effect), these let you add a color or luminosity bias to your gens. Similar to post-production in photoshop, but since the model can already "see" this during generation it can adjust the scene to match. Like by removing light sources in a dark scene for example, or adding a sunset glow if you bias it towards orange, etc.

Test pic is genned on Lustify, same prompt and seed except different Vectorscope settings.

damn, effort posts take too long

I hope my crippling mommy fetish isn't obvious

>>8946222
I was trying out BigLove XL4 and BigLust 1.7, BigLove is more controllable with keywords to make the scenes darker while BigLust practically turns the subject white and the background pure black.
Tried Vectorscope with -2 contrast and it worked pretty well. probably will stick with BigLove, BigLust fails to follow certain prompts like flashing and putting on/taking off clothes.

>>8946212
Mostly I wanted to generate cyborgs, but the guy is usually the one who ends up with tattoos.

>>8946297
Vermeil? She looks softer than I imagined.

>>8946253
If it weren't how could I enjoy this

>>8946615
Man of spectacular taste located

>>8946540
yes its her.

>>8946297
It is clear that summoning demons is the only way forward for modern men

>>8945897
you can also img2img from a black image at high denoising. This also works with other colors or gradients if you want to tint your photos a certain way - picrel is the Floral Shoppe album cover at .9

>>8946384
It's cool. Thanks for trying anyway. I don't want to be that guy, but try to make sure only the women have dark skin if that generates. No one wants another blacked thread to propagate.

>>8946755
Oh come on, that guy's face doesn't look remotely black.

>>8946783
>Oh come on, that guy's face doesn't look remotely black.
A brown man is brown man. It's still an invitation to the rabbi. Also, very nice. I have a thing for judy over panam even though they purposefully blocked out her ability to be romanced by men. So fucked up.

>>8946259
>BigLust 1.7
You used the wrong one. Only 1.6 matters.

>>8946755
Oh look, another schizo

>>8947027
you must not have been here when the moderators had their IP locations and government affiliations leaked.

>>8947078
That's so nice

>>8947260
Diamond inducing. Moar please.

>>8947472
fantastic

Is this thread suitable to semirealistic paintings as well?

>>8948300
Yes.

>>8948308
Ok then. Today I feel like resurrecting an old artstyle I used to enjoy. Maybe some of you will also like it.

>>8944438
nice

>>8944697
Is that Delia or Deidre?

This thread here is fun with the idea of creating cartoon characters from photorealistic checkpoints. And also a fun challenge of prompting their features without Loras, to see like how some characters would look like in say, Sambase.

Just a disclaimer ahead the eyes might startle some of you guys. Apparently in the anime this character has pink eyes, and I just put some prompts of her attributes into Sambase, but the eyes are naturally from the dataset (no eye prompts except only large eyes as I like my genned girls with big eyes). But my attempted version from this checkpoint turned out like this xD

>>8948810
They'd look a lot better with an inpaint pass or adetailer. Same for >>8948721

>>8948842
Ok. For inpainting is it the one within ControlNet? For inpaint pass what roughly is recommended that can be improved? And where to find adetailer?
For the time being like yeah I've been sending out or posting just base gens, and haven't yet moved to any enhancements that would improve on the base gen.
And btw, where do I find the regional prompter?

>>8944432 (OP)
Do you need specialized setups to make this hyper detailed almost more realistic than real life images? Or is this just something you achieve through tinkering with options that doesn’t require anything more complicated than simple cartoon stuff? Do you need more powerful hardware thr more detailed something is? Or does it simply just take longer? I find these really cool to look at occasionally despite finding realism to be sometimes boring in art.

>>8946253
This is amazing and a good aesthetic but I think you could push the MILF status slightly futher by 5 or 10% for true perfection.

>>8949011
>>8949011
Glad you like them!

>>8948617
>creating cartoon characters from photorealistic checkpoints
Oh some of us have been doing that for quite a while now

Does anyone have the prompt and setup metadata for SparklingSenpai's gens?
Can't find anything anywhere

>>8949099
I guess it was beginner's excitement hahaha. But I will probably be careful or cut down on oversharing, especially just base gens as I'm slow and doing this casually. Btw bumping these questions >>8948870

Requesting some realistic gens of Kyoka Jiro, cute flat tomboy from My Hero Academia.

>>8948870
How to use these tools depends on which UI you are using.

Anyway, the AI struggles to produce good detail at low resolutions, especially for objects small or far from the camera. A photograph with not enough resolution would be blurry, a drawing would force the artist to simplify things, and an AI gen will have weird distortions usually in the eyes, mouth, fingernails, zippers, jewelry, etc. Like how nezuko's eyes seem to have a sharp edge halfway through, and the random dot of orange above her right eye,
or the light reflection in Marin's eye being three disconnected reflections instead.

Upscaling is a brute force way to fix these across the entire image, but it takes a lot of time and you might not be able to reach high enough resolution. Inpainting does a crop+upscale of only the selected area which takes seconds. Adetailer stands for after-detailer, it is simply an automated tool for inpainting faces immediately after generation. Because this resolution problem very commonly affects faces and especially eyes.

>>8949010
nta but all hardware produces the same output, as long as it doesn't crash outright. It takes longer the higher the image resolution is, but that doesn't necessarily mean it's more detailed. I would guess it's either his prompt or a lora specifically for this aesthetic. Also being a close-up shot helps with the resolution detail issue.

>>8949413
I'm using Forge. I'm trying to find Adetailer and also the regional prompter if that helps with multiple characters. Anyway for inpainting I can use the inbuilt one in ControlNet? Like is it the same technique that you used for Misaki, so you can either fix or add effects to eyes or face right? Inpainting is usually after already producing a base gen?

>>8949360
For today I'm feeling generous but bear in mind that this was a bit rushed and probably low effort. I'm no wizard and this is one off. Experienced genners would just find a character Lora in civitai and shortcut the efforts instead of manually prompting the character's attributes with the help of referring to Danbooru like I've been practicing. But beginners always willfully start from somewhere, like I got Forge, tried out checkpoints you're confident with and decide yourself what you wish to do....
Sorry TLDR but here are the below anon.

>>8948153
This one is interesting as I haven't attempted more than like a 1boy 1girl. Is this 4girls and trying to keep the prompts simple?

>>8949550
>>8949551
>>8949552
Actually a fantastic start, thanks anon.

>>8949544
>Inpainting is usually after already producing a base gen?
Yes, it works with existing images.
>for inpainting I can use the inbuilt one in ControlNet
You should be able to find inpainting guides for Forge easily. I don't use Forge, but from what I've heard anons mention it should be the img2img tab I think? Controlnet is a different tool altogether. One of its uses is aiding with inpainting in specific cases, which I guess is why you saw inpainting mentioned there.
>regional prompter if that helps with multiple characters
I believe it's called Forge Couple over there and has to be downloaded as an addon. It lets you define some areas in the image and give each of them a separate prompt.

>>8949360
She is a weird anomaly, works very poorly despite being mainstream with a huge number of images.

Bonus, trying to inpaint "horse dildo" with Lustify
wish the non-anime SDXL models weren't all so stupid

>>8949130
He makes money off of it, why do you think he would ever share? You can train a lora on his images if you really need the style.

>>8949585
hilarious. i think you can fix this if you bring down the denoise a little bit.

>>8949413
great explanation

>>8949010
nta, really it's just the right checkpoint, loras, and doing some inpainting

>>8949066
bone apple the teeth

>>8948610
Your pick.

>>8949920
I'm gonna say left and >>8948498 are Delia
Right and first pic are Deidre
That's my final answer

>>8949558
See catbox. I've done a lot of experimenting with multiple women/lineup and IMO what >>8949413 said applies extra to pictures with multiple women. You usually will need adetailer for a second pass or to do the women individually with a regional prompter to avoid deformities and sameface. In general, when proompting 1girls, I get an aesthetically pleasing image ~75-90% of the time. That goes way down when I try to include more than one girl, especially if the picture is zoomed out.

Flux and Chroma are better with this than 1.5/XL but it's still not perfect. I haven't tried adetailer with them but I assume it's basically the same

>>8950447
Are you the same anon who used to generate women of various ethnicities back in the SD1.4 days?

One anon granted me 2 wishes and I now have 2 LoRAs of girls I have been collecting for a while.

>>8950475
maybe, I remember I tried to mess around with nationality wildcards at one point, i posted some real slop with lineups and controlnet lol

>>8944569
what model is that? Id love to be able to create realistic pictures of celebrities.

>>8950503
As long as your specs are above an NVIDIA GPU with 4gb vram.
The best one for me is Big Lust v1.7 with DMD2 LoRA (it increases speed by making generation with 4-8 steps possible).
The celebrity LoRAs are a bit harder to make but I'm sure anyone on /b/ will be up to help.

>>8950537
That one's really nice. I wish there was a better way to bridge models that recognize anime characters with models that look like actual photos.

Crap, think I missed the tailbone.

>>8950969
The cartoony side: >>>/aco/csdg

>>8950977
Thanks

>>8951393
thats amazing

Chubbies <3

>>8951523
>>8951568
i assume this is someone you know IRL and i think it's kinda funny how many people have a plus size latina in their lives who they would gladly creampie

>>8951586
Yep, good to fuck but great to cuddle.

I really shouldn't have access to ai

Going full yandere with these.

>>8944432 (OP)
I struggle to find realistic Loras for Illustrious. Not interested in real persons, just fake girls next door kind of thing. I use them to get somewhat consistent faces. Animu/comic Loras don't have distinct facial features / flaws (except for hair).Do you guys know any? I've only really looked on civitai so far. Things like pic related (sadly it's SD1.5, not illustrious).

>>8952443
You need realistic checkpoints more than loras.

>>8952443
As >>8952448 says, look for a realistic checkpoint in civitai. Maybe one by Stable Yogi will work for you?

>>8952448
>>8952463
I have a few checkpoints that I like, including a yogi's one. But I find the faces often samey, almost too perfect and I struggle to make imperfect faces with defined characteristics. I tried with a few loras and I get the expected results, I'm just looking for more. Also I'm an esl prompting noob so that doesn't help. I often struggle to find the good keywords, especially for face description. I use artist/cosmetic references charts, like for nose shapes and all, but it doesn't seem to really work most of the time. I'm really after that average girl next door look.

>>8952535
So from what I gather Illustrious based models use Danbooru tags. But since those tags are made for animu, there really isn't any tags to describe facial features because well anime faces don't really have any. To it's kind of useless to try. Is there any model (type) that supports faces prompt?

>>8952673
Well, they do have some facial features. Like freckles, or wrinkles. But even if a tag exists on danbooru's tag wiki, it might not translate well to photorealism.

Base SDXL itself handles facial detail well enough, and might work for inpainting. As for the nsfw merges and finetunes we use here, it always depends on the author's dataset. Whether he was consistent with tagging, whether he even tagged faces in such detail in the first place. Which words and phrases he used. There's no way to tell other than manually testing stuff.

>>8952722
Yes I understand thank you. I'm already amazed at what I can produce when I just started to look into this this week. Pretty crazy stuff. I must check this WAN animation stuff too if my PC can handle it.

>>8952689
A very interesting set.

>>8953730
I love your Dee Dees.

>>8953734
Thanks.

>>8953738
You're welcome and thank you.

>>8952096
>>8951826
These are really neat, are these sora?

aqua-sama
i thought it was more "realistic" so it could go here too :D
- illustrij model
- some detailers
- aqua lora
- gen > upscale > facedetailer

>>8950520
>
>>8951117
Is this meant to be Ruby from RWBY?

>>8950475
More of this cutie redhead doing everyday things in the nude please.

>>8954242
Yeah, some models do it better than others.
These are from a while ago, though.

>>8954032
Yesssss

I finally got video "down" after learning a bit more about Framepack Studio. That said, it still feels like I could be doing better. It takes like 30 minutes to go from initial gen to final moving video. Have some mommyslop while I try to downscale this webm

Still trying to find a model that 'feels right'

>>8954926
asianmom slop

I stumbled upon this https://civitai.com/models/1608870/dmd2-speed-lora-sdxl-pony-illustrious
In the past a pic took about 20 minutes to gen (potato basic laptop). I haven't tried this lora yet. According to the civitai page, they have their list of settings which seems to be reduced CFG, steps, different samplers. My concern is I'm so used to the original settings when genning in Forge with their checkpoints and specific setting, so which settings should I then follow? Like might it affect my gens compared to before? But this lora speeds up gens right, and how to use it?

>>8953875
Krea ai. Made a lora a long time ago when they were using flux, but these are with their Krea1 model.

Don't know how nsfw it can go, but for me it's more the implications and aesthetics that matter for now. Gosh, I wish this tech could develop faster. I want to plug my brain into the Matrix for an hour and have a stressed out angry teacher Nimoy lock me in his office and just go wild on me.

>>8955444
>>8955437
Hot

>>8955444
>angry teacher Nimoy lock me in his office and just go wild on me
Oh, my.

butt

>>8955444
If you're having trouble prompting NSFW content, you could try inpainting whatever you're interested in onto Spock.
https://www.youtube.com/watch?v=64tLb79YRbI

Made this one accidentally, love the nip slip

>>8955803
How'd you get upward tilted tittes

>>8955761
>>8955803
>>8956545
Really nice work.
>>8957469
Upturned breasts, either by token or LoRA.

I was told this might be appreciated here

>>8957833
interesting combination, it's like a real photo with a genshinface. I think the head is a bit too large
did you gen this in one go?

>>8957857
yeah, in one go

>>8956891
more of her, I love this bimbo korean

>>8957469
Already deleted that one.
Made it a long time ago.

Different models have different triggers.
perky breasts works for a bunch of illustrious models.
There are a couple loras for swooping breasts and steep slope tiddies.

Does anyone use nudify lora for flux kontext?

>>8957934
Are you using PornMaster-Pro with the realistic skin LoRA?

>>8958788
I think that was just PonyRealism.
I cant quite get the Illustrious models to look the same, but they are so much more flexible otherwise.
Pic also unrelated.

no razors in the post-apocalypse

>>8958207
does that even work?

>>8959597
Trimmer would be enough

>>8959667
even scissors would help at this point

IN THE NAME OF THE PLANET EARTH, I CUM IN PEACE!

>>8959692
is that a guy or a coral formation?

>>8959793
Maybe both? I didn't even want the guy to be an alien, I just wanted them to be fucking on an alien planet.

>>8956891
Cool. Inpainting using what tool?

Man I guess I'm getting laid tonight jeez

woah

I honestly don't think he would've wanted fake images like these floating around but I just can't help myself, ai is too addictive

>Finally tried my hand at making some AI porn yesterday and thought I made some pretty good stuff
>Look through all my saved images today and realize 95% of them have fucked up distracting eyes
Here's one that is not so fucked up cause I just copied a prompt

>>8960194
Because her eyes are relatively close-up. In full body shots you have to inpaint the face pretty much every time.

>>8960542
>That pubic hair peeking...

"Ho honey, why are you looking at me like that ?"

>>8960557
that's the best part

>>8944432 (OP)
What is the best age slider?

>>8960194
You messed up when you gave me this

>>8960542
>>8960736
kino taste

Some photorealistic models have difficulty getting simple bangs or hair between eyes even with heavy weight. Do you know a workaround that?

>>8944702
Eww..
>>8946207
Are those dirty garden hoses on her head? wtf? Disgusting

>>8953788
Hot, but what is wrong with her hair, what are those supposed to be? Is this a style women actually wear? And why does her skin look like someone smeared poop all over her?

>>8961465
Are you using models that recognize Danbooru tags? They should if they're Illustrious, Noob or Pony-based.

>>8961487
yeah, three of them, but hair between eyes tag is something they don't recognize.

>>8959897
I use tensor.art for my inpainting needs since I'm a poorfag but it gets the job done.
>>8957931
sure
https://www.youtube.com/watch?v=oQR92qFBlQA

>>8961489
You might have to use the NegPiP extension and add something like (forehead:-1.5) to the prompt.

>>8961520
Thanks, didn't know that extension exists.

>>8961520
I know many of them are overfit on just a few faces. But still, that seems way overkill for such a simple tag. Not to mention one with over a million hits.

>>8961574
Hair between eyes isn't that common of a hairstyle in the west these days. Even if you give the tag 1.5 positive weight, it only generates correctly about half the time.

Just started a few days ago and I'm trying to get a handle of the tools. (Thanks for the great OP.)
This is my most ambitious image yet, and I'm running into a rather interesting problem. Evidently, the quality of the characters and prompt seems to degrade the further they get from the 'camera'. I think some of this can be fixed with in-painting (haven't tried it yet), but there's gotta be a better way, right?

>>8961583
Yes, you can use ADetailer with a face recognition model. It automates the inpainting process for you.

>>8961582
Sure, but if we're talking Danbooru tags then it's pulling from anime knowledge and not the real world. I can even do weird stuff like short_hair_with_long_locks, or drill_hair.

Guess it depends on how much knowledge was lost in converting that specific checkpoint to photorealism. It's why I shilled Sambase in the last thread.

>>8961591
Sam Base isn't realistic enough for me. I've been using Realism Illustrious.

>>8961591
I think a good way to test this is asking for a less well-known anime character. See how much of her hair color, hairstyle, eye color, and clothes it can give you. https://files.catbox.moe/i7w44u.jpg

>>8961583 check >>8949413
For >>8961587 the ComfyUI implementation of Adetailer is called FaceDetailer, from ImpactPack custom nodes.

You can also upscale the whole image, then do a second AI pass on it at low denoise.

>>8961593
Yeah seems like it has to be a compromise, the closer a model gets to actual photographs the more knowledge it loses along the way.

Could also setup the scene with one, then img2img with another to fix the look. But that's maybe too much work.

>>8961587
>>8961595
Thanks, I'll investigate!

Hello, I came across DMD2 speed lora https://civitai.com/models/1608870/dmd2-speed-lora-sdxl-pony-illustrious
Haven't tested this out. I was previously taking several minutes to gen one pic in Forge with let's say an Illustrious checkpoint. I am also used to the settings of the Illustrious model i.e. CFG 5, steps 30, sampler Euler A. So the civitai page recommends different settings, lower CFG, steps and different sampler. But activating this lora is like any other lora, and its reduces genning time? Does it mean any of my future gens should be this lower CFG and steps? Anyway just asking as I haven't tested it yet.

>>8961593
Man I tried Realism Illustrious by Yogi and you're right, it's completely lobotomized. I wonder what was the point of basing it on illustrious if he's just gonna burn away everything the model used to know.

>>8961618
Yogi is a hack. You can have a hundred tokens of SFW woman walking down the street and then you add large breasts and she immediately goes topless, meaning he didn't tag breast size for women wearing clothes. Wouldn't surprise me if he also didn't tag hairstyles beyond short/long and color. Not a matter of losing knowledge if you explicitly teach the AI that hairstyles don't matter.

Here's a challenge, try to make the girl look anywhere other than the camera.

>>8961610
sigh, this general is useless. normal gen vs 4 step 1 cfg with euler a/beta
which is which should be obvious.
>But activating this lora is like any other lora
yes

>>8961667
Useless for questions you can get an answer to yourself in two minutes.
>I haven't tested it yet
Why not?

>>8961667
>sigh, this general is useless.
Lol

Thanks Anon. At least I have an idea now :)

So, I'm trying to have a voluptuous woman that's more on the wider hip side of things but whenever I try it out the outcome is always fat tummy. I'm using pixverse and I'm basically wanting shortstack proportions. Can I get some help?
The idea is to get this body type
>Pic related

I know this gen isn't listed but any help would be appreciated, thx

>>8961711
Hourglass figure. Maybe heroic proportions?
Remember to put hourglass in the neg or things can get funny.

>>8961730
How do you mean in the neg? Curious because I'm not familiar with that. Things are looking funny already TT.TT

>>8961731
Negative prompt. Put anything there you don´t want on the image, like chubby, or fat.

>>8961489
Are you using the actual tags for it or making up your own keywords?

what's a good start pic/end pic progression prompt for BE on pixverse?

>>8961859
>>8961861
>>8961862
Nice

>>8961691
>>8961693
Could you share the metadata for either of these? Pretty nice style.

Post a link to the tampermonkey thread

Is there anyway to freefag kling any more? I made a new gmail account and it always tells me the slots are filled.

anything better than wan 2.2?

>>8961763

It depends on models. Some will lose danbooru tags when converting to photorealism-focus checkpoint.

>>8961465
inpainting

>>8962174
Githyanki babe <3 Laezel has smaller breasts though

>>8961691
>>8961693
https://files.catbox.moe/13tlib.png
https://files.catbox.moe/fj31pr.png

>>8962642
thanks!