/vt/ - Virtual Youtubers

Request Anchor

roboco waiting to be bukkaked
https://files.catbox.moe/ina5l9.png

>>46989720
I volunteer, but I won't take responsibility if she short circuits!

>>46989773
This Ina looks really fucking cool

>>46989604
https://youtu.be/-Grs8KxXO90
https://youtu.be/5JiXqzd0Xdc
>Senchou stealing another woman's boyfriend
>Yandere Senchou reducing the man's hips to powder
>Smug pregnant Senchou
>Mature Senchou with her future children

>>46989604
Requesting big booty hologamers. Talking built like a glass part pointing down lightbulb thicc

botan was gacha'd from the wheel next https://files.catbox.moe/ccsq7t.png

>>46989604
hime cut mio in a sundress

You gonna make those prompts of yours {{{bigger}}}?

>>46989604
Botan machine gunning down russian soldiers

>>46990149
Highjacking because i need to see botan like this

>>46989604
>Gooba date rape
>Gooba checking a pregnancy test
>Gooba pregnant with enough sharts to fill a baseball team
>Proposing to Gooba (preg or not)
Skullgirls or Don't Starve loras

>>46989604
Skyler Reach in a chemise

Quick Addendum/Repost to my original merge Post last thread.

>What the fuck is a Bayesian Merge? And why the hell are you so excited about it?
Well, it's quite simple, Bayesian Merging takes everything that is so great about MBW Merging and does it for you. And unlike other Auto MBW mergers, this isn't a brute force algorithm, but rather does Black Box Optimization on the merge. When combined with a scorer, you can achieve results that would have taken days to produce and test in only a couple hours. And I think the results speak for themselves.
>Well that sounds great and all, but I produce perfectly fine results with standard merging.
Well, MBW Merging is for merging two models without losing/diluting the concepts from either. The BBW model is an almost perfect candidate for MBW Merging, because with any sort of standard merge, you lose/water down the main concept of the model due to the fact that fat anatomy is so rare in the NAI dataset, and in general there is a very small dataset of art for the concept, which makes it easily overwhelmed when merged with another model. The fact that I didn't have to screw around with individual Block weights and continually test everything manually is a massive plus, since it saves time.

All in all, Bayesian Merging was absolutely worth the trouble it took to fix (it's now fixed on main as well so you should be able to use it), and configure. I'm curious to hear some feedback on how it performs since I went to some pretty big lengths to make a pretty large payload set to cover a number of holo chuubas.

Here's the original post:
I finished the final version of the bbw_hll3 merge.
Test Images from all 13 payloads. (Some of these have Dynamic CFG, some of these don't, all of them have been through Hires fix. These prompts were all run on 20 steps DPM++ 2M Karras( with most having 7.5 cfg) when scored) There is 1 non-fat prompt, where I just prompted for a character with no modifiers, and one prompt that should result in an image that isn't fat focused. (Polka and Noel)

Test Images from all 13 payloads. (Some of these have Dynamic CFG, some of these don't, all of them have been through Hires fix. These prompts were all run on 20 steps DPM++ 2M Karras( with most having 7.5 cfg) when scored) There is 1 non-fat prompt, where I just prompted for a character with no modifiers, and one prompt that should result in an image that isn't fat focused.(Polka and Noel) Additionally, both the OP Image and the request Anchor were generated from this model.

Suisei: https://files.catbox.moe/izd9b9.png
Watame: https://files.catbox.moe/zj0uj7.png
Polka 1: https://files.catbox.moe/9yxuo0.png
Polka 2: https://files.catbox.moe/i8v713.png
Polka 3: https://files.catbox.moe/km66fr.png
Noel: https://files.catbox.moe/7ari39.png
Mumei: https://files.catbox.moe/jo92in.png
Mio: https://files.catbox.moe/cloew1.png
Lui 1: https://files.catbox.moe/bjeein.png
Lui 2: https://files.catbox.moe/u6xa1y.png
Lui 3: https://files.catbox.moe/poink3.png
Lui 4: https://files.catbox.moe/llvbl9.png
Laplus: https://litter.catbox.moe/gi73mt.png
Kronii 1: https://files.catbox.moe/gfqfo1.png
Kronii 2: https://files.catbox.moe/3ci6lt.png (using DPM++ 2s a Karras at 41 steps)
Irys 1: https://files.catbox.moe/gnrtrb.png
Irys 2: https://files.catbox.moe/dqnob8.png
Irys 3: https://files.catbox.moe/aaqmi1.png
Ina: https://files.catbox.moe/80rcfc.png
Haato 1: https://files.catbox.moe/e5k7is.png
Haato 2: https://files.catbox.moe/1unn0s.png (detailed background added to tags)
Ayame: https://files.catbox.moe/pknk3a.png

Model Download is here: https://huggingface.co/mariaWitch/ExperimentalBiggerMerges/resolve/main/bbw_hll3.1%20Gods%20Amongst%20Men%20Edition.safetensors

>>46990149

>>46990149
>>46990622
KEK

>>46990386
Nice background, is it from the model(mix) used or a specific set of prompts?

>>46983070
Just the war criminal rabbit burying the Demon Core

>>46990700
NO PEKORA DON'T DO IT

>>46990622
Needed that laugh, thanks anon

Quick Addendum/Repost to my original merge Post last thread.

>What the fuck is a Bayesian Merge? And why the hell are you so excited about it?
Well, it's quite simple, Bayesian Merging takes everything that is so great about MBW Merging and does it for you. And unlike other Auto MBW mergers, this isn't a brute force algorithm, but rather does Black Box Optimization on the merge. When combined with a scorer, you can achieve results that would have taken days to produce and test in only a couple hours. And I think the results speak for themselves.
>Well that sounds great and all, but I produce perfectly fine results with standard merging.
Well, MBW Merging is for merging two models without losing/diluting the concepts from either. The BBW model is an almost perfect candidate for MBW Merging, because with any sort of standard merge, you lose/water down the main concept of the model due to the fact that fat anatomy is so rare in the NAI dataset, and in general there is a very small dataset of art for the concept, which makes it easily overwhelmed when merged with another model. The fact that I didn't have to screw around with individual Block weights and continually test everything manually is a massive plus, since it saves time.

All in all, Bayesian Merging was absolutely worth the trouble it took to fix (it's now fixed on main as well so you should be able to use it), and configure. I'm curious to hear some feedback on how it performs since I went to some pretty big lengths to make a pretty large payload set to cover a number of holo chuubas.

Here's the original post:
I finished the final version of the bbw_hll3 merge.
Test Images from all 13 payloads. (Some of these have Dynamic CFG, some of these don't, all of them have been through Hires fix. These prompts were all run on 20 steps DPM++ 2M Karras( with most having 7.5 cfg) when scored) There is 1 non-fat prompt, where I just prompted for a character with no modifiers, and one prompt that should result in an image that isn't fat focused.(Polka and Noel) Additionally, both the OP Image and the request Anchor were generated from this model.

Suisei: https://files.catbox.moe/izd9b9.png
Watame: https://files.catbox.moe/zj0uj7.png
Polka 1: https://files.catbox.moe/9yxuo0.png
Polka 2: https://files.catbox.moe/i8v713.png
Polka 3: https://files.catbox.moe/km66fr.png
Noel: https://files.catbox.moe/7ari39.png
Mumei: https://files.catbox.moe/jo92in.png
Mio: https://files.catbox.moe/cloew1.png
Lui 1: https://files.catbox.moe/bjeein.png
Lui 2: https://files.catbox.moe/u6xa1y.png
Lui 3: https://files.catbox.moe/poink3.png
Lui 4: https://files.catbox.moe/llvbl9.png
Laplus: https://litter.catbox.moe/gi73mt.png
Kronii 1: https://files.catbox.moe/gfqfo1.png
Kronii 2: https://files.catbox.moe/3ci6lt.png (using DPM++ 2s a Karras at 41 steps)
Irys 1: https://files.catbox.moe/gnrtrb.png
Irys 2: https://files.catbox.moe/dqnob8.png
Irys 3: https://files.catbox.moe/aaqmi1.png
Ina: https://files.catbox.moe/80rcfc.png
Haato 1: https://files.catbox.moe/e5k7is.png
Haato 2: https://files.catbox.moe/1unn0s.png (detailed background added to tags)
Ayame: https://files.catbox.moe/pknk3a.png

Model Download is here: https://huggingface.co/mariaWitch/ExperimentalBiggerMerges/resolve/main/bbw_hll3.1%20Gods%20Amongst%20Men%20Edition.safetensors

>>46990691
Not at all, I think this comes from the techpunk mask LoRA I'm using. It automatically inserts all kinds of cyberpunk and weaponry into the image (unprompted): https://litter.catbox.moe/s799lt.jpg

>>46990414
Wide load Botan

>>46990804
FUCK.
I'm not deleting this again. I'm attaching the image I wanted to attach to this post here.

>>46990804
>Model Download is here: https://huggingface.co/mariaWitch/ExperimentalBiggerMerges/resolve/main/bbw_hll3.1%20Gods%20Amongst%20Men%20Edition.safetensors

Is that the only link I need to add/update for the Colab?

and for rentry anon, is it the only one they need to add a hash for?

>>46990881
Yes. That is the only one to add. All of my non-standard/unreproducible merges are going there.
Something weird with Supermerger makes it impossible to achieve the same model hash as the model generated by Bayesian merger even when the mbw values are all the same.

>>46990398
Kinda hard to force a hime cut on her

I arrive in here and immediately find out about a new model-merging strategy. I really should show up here more often.

>>46991006
that's rentry anon's broblemo if he hasn't died again. you already gave me a /resolve/ URL so I don't have to do shit kek
>new repository
fuckin...
was about to say it was already up, but a 404 link is technically already up. 2 minutes.

>>46990852
Wide botan

>>46990398
ditto

>>46989837
but short circuiting a robot is the best part

>>46990700
I expected grenades, but Pekora goes hard. God help who ever finds this "egg"
Thanks anon!!

>>46990414
I'm eventually going to end up with two house-sized breasts on a normal-sized Mio

well are you gonna give your pet succubus her food? https://files.catbox.moe/evezc5.png

>>46991164
MBW Merging or Bayesian Merging?

>>46991576
Bayesian. I've heard of automatic MBW weighting before, I have Silicon downloaded. But I haven't seen anyone actually use this stuff other than the creators.

>>46991409
Box?

>>46991739

https://files.catbox.moe/y6hb8n.png
https://files.catbox.moe/jpq3q9.png

>>46987639
>>46976462
original fox belly requester here. i ended up falling asleep before you came back but thank you so much!

>>46985592
>>46986714
beyond based

>>46991831

>>46991686
https://files.catbox.moe/iv0e32.png
>>46991739
She looks so good in black lingerie

>>46991642
Yeah. I actually think it creates better merges than the Auto-MBW merger. The only problem is that the model hashes are not reproducible on the only merger extension that can replicate the kinds of merges that Bayesian Merge can do. Which sucks, but in theory, the MBWs should lead to similar results.

>>46991885

>>46991975
>>46991642
And honestly the additional information it gives you even after the merge completes is pretty awesome.

>>46992006

>>46991975
>>46992178
I'll be sure to check it out, seems pretty cool. The most complexity I usually do with merging is adding 0.3 of 20 different loras in. And it not having reproducible hashes is fine. I have multiple mixes I completely lost the recipe for, so what's one more.

>>46989604
>Wawa moaning in delight while an unamused Officer Shuba claps her on the ass
https://youtube.com/shorts/JH1AropcPeo?feature=share

Good afternoon, /vtai/

>>46955512
Here you go
https://files.catbox.moe/o4q867.png
>>46989331
>>46988210
>>46987871
>>46987554
>>46987316
These remind me of Radagon from Elden Ring, pretty neat.

AVE
https://files.catbox.moe/1hc162.png
https://files.catbox.moe/go4ppt.png

>>46992324
I wonder if i can beat them at table tennis

>>46992385
Can i get a box? thanks in advance

Good night

>that tiny glass on the left side of the image
KINO

>>46992603
https://litter.catbox.moe/3jw171.png

>>46992618
Implied amazoness/mini gts is hot shit.

>>46990804
>Rentry updated
I added it, very sloppy because I don't even know where to get bbw2x5.0+1.0 nor do I want to think of a format so I pretty much just pasted your instructions and linked the original post

HLL4 Bayesian Merge coming, although it's going to be a bit more complex.

Double weighted difference for those who don't know is (1 - alpha) * model_a + alpha * ((1 - beta) * (model_b - model_c) + beta * (model_d - model_e)

Mama Sana

>>46992913
You can find it in the same repo: https://huggingface.co/mariaWitch/ExperimentalBiggerMerges/tree/main/Archive

>>46989604
Tired of the coom? Want some new ideas? Come on down to the creative promt shop! Try some of these on for size!
(Not requests, just for fun!)

1. Sitting by the water dipping her toes in (could be on a river/pond bank, or off of a dock. I'm not thinking of a beach.)
2. Lolified vtuber doing a coloring book
3. Tenma (or your choice) in an Irish cap throwing back a pint (preferably in a bar)

Have fun!

>>46992996
Will add it, thanks

>>46989604
Requesting more chubbas wearing Sesshouin Kiara swimsuit:
Selen
Hisui short hair (Niji) (if there's a LoRA of her)
Reimu
And thank you for this kind anon who made an extra Botan >>46968798 really sexo!

>>46993053
> Tired of the coom?
No

>>46993120
Are you looking for some new ideas then?

>>46992993

>>46989604
>Gooba calling (You) stinky
>Instant loss sniffing (You)r wet dick

>>46993174
sure, actually

>>46990804
Well you seem to know what you're talking about. I D/L MBW a days ago but haven't used it. If what you're saying is that it makes it so you can merge 2 models without them clobbering their tags then what are some preset that give decent results or at least a good starting point.

Let's say I want my holo model to take some cyber robot concepts from some model off Civitai without my chuubas tags getting wiped, which preset should I pick?

What are the IN and OUT sliders mean? I tried reading the manual but shit is like a textbook, it says some of them control art concepts, lighting, shapes etc..

What preset did you use? I don't mind doing some brute force but I'd like a good starting point.

>>46990350
>>46991739
>>46992993
>>46991884
I love seeing so much Baby for Easter!

>>46992618
Based

>>46993283
this is cool

>>46993213

>>46993740

>>46992723
that uniform is kinda lewd, moom

>>46991542
I think you should make one of her getting her meal

Tell me about Koyori, what makes her such a sexy hag?

>>46993839
how does she get it? just placed in front of her or directly put in?

>>46993295
>without them clobbering their tags
I wouldn't say that you are merging tags per say, since that would imply that you are merging the clip encoder rather than the model itself. You are merging the UNET Blocks, which is incredibly abstract. (Although if you also want to fuck around with merging the clip encoder, then mucking around with base_alpha will control how much of model_Bs Clip encoder is merged with Model A's.)
>Let's say I want my holo model to take some cyber robot concepts from some model off Civitai without my chuubas tags getting wiped, which preset should I pick?
There isn't one preset to rule them all. You can use a preset as a starting point and go from there.
>What are the IN and OUT sliders mean? I tried reading the manual but shit is like a textbook, it says some of them control art concepts, lighting, shapes etc..
Okay, so the IN sliders deal with the IN blocks in your UNET, these downsample the image while adding concepts until it reaches MID, at which point the UNET starts upscaling the image and adding concepts again. The IN and OUT sliders do the same thing, they just refer to different locations in the UNET. A good rule of thumb is that concepts, and compositions are going to be primarily in the IN Blocks, while they control texture and colors secondarily, the OUT Blocks are primarily responsible for Colors and textures, with Concepts and composition coming secondarily. Pic Related gives a ~~~~VERY~~~~ approximate description of what could be in which block, however every model is different and concepts may be else where due to different weights of the prompts and tensor deviations.
>What preset did you use? I don't mind doing some brute force but I'd like a good starting point.
I didn't use any presets, I used a black box optimizer in order to find good merge values. However, if you look here: >>46992178 , you should be able to gleam some critical information about what might be where for the HLL model at the very least. The darkest red areas are areas where the HLL blocks completely replace the BBW blocks. If you are merging an HLL model into another model keep a close eye (which is to say, make these blocks closer to 1) on IN00, IN01, IN04,IN05,IN06,IN08, and OUT02, OUT07,OUT09, and OUT10. As the anatomy concept you are talking about falls in the areas that are more blue in the post I mentioned above. I would personally start with the flat 25, and just screw around from there if you are going to do it manually.

>>46994065
I bet she's hungry, directly putting it in seems fitting

>>46993053
Fauna relaxing at the canal after a good day's work

Not quite what I was going for but still one of the best generations I've gotten so far I feel.

Bonus open top from rng too:
https://files.catbox.moe/qumozb.png

>>46990149

>>46993901
That tummy is definitely helping. Catbox please?

What's the deal with AI and fingernails?

>>46994252
gentiles are so difficult to do https://files.catbox.moe/24l51l.png

>>46994621
https://files.catbox.moe/y8yg31.png

>>46994667
Make your life easier and just put "nail polish" in negatives

>>46994719
Thank you.

>>46991093
>>46991313

thats quite cute!!!

>>46993697

>>46989604
>Platinum Kiwawa
https://youtube.com/shorts/pdp0-FYfNuk?feature=share

>>46994518
Kino, would the gods allow {{bigger}}?

>>46994692
Nice, love the blowjob face, good work. Also, seems you didn't have glasses in the tags, but it generated some anyway lol

>>46993053
Ina doing coloring books inside her cell

>>46993053

>>46995188
NTA but cute! Could you catbox that please?

>>46995281
Beautiful. What background tags are you using? I've been defaulting to "wadding pool"

>>46991884
are you the anon that did the collection of fattest gamers a few threads ago in a 7z archive? If so could you re-up that?

>>46995390
My only background tags were at the beach, sitting on dock, and feet in water. It genned the rest.

>>46995286
Here you go: https://litter.catbox.moe/scdabc.jpg

>bbwm-bbw2x5.0+1.0v-half-hll4-p3-ep8-final-pruned-pruned-fp16-Silicon29-dark-final-pruned-pruned-fp16-it_1.safetensors
Great name. I love it.

>>46995602
Thank you.

>>46995188
>>46995602
>>46989604
>Lolina Fingerpainting classical art

>>46995391
https://litter.catbox.moe/xvdre3.7z
These?

>>46995758
Got more like that Mio spoiler?

Goddamnit anyhentai just updated, maybe it'll be less finicky.

>>46996031
Sadly no, but I can make more later.

Is Silicon29 a WD model? It uses a WD VAE, but seems to produce output incredibly similar to Anything v4.5 in their comparisons.

>>46996356
Beyond based, make em files catbox in case i pass out from how good they are

>>46993053
Prepare your kokoros
https://files.catbox.moe/ox0ems.png
https://files.catbox.moe/a9fzrk.png
https://files.catbox.moe/sqrkd8.png
https://files.catbox.moe/n1xfma.png

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

This machine looks familiar

>>46996747
INA NO

>>46996747
I KNEW IT WAS COMING, I COULD SMELL IT ON THE HORIZON >:U

>>46996747
Don't you dare, anon... D:

>>46990850
So if you inpaint these and remove the tokens describing the character, it would spit out just the background, that can be photobashed with other stuff?

>>46992324
Goodbye /vtai/
Taking that month-long vacation.
>>46992603
>I wonder if i can beat them at table tennis
Choose your partner wisely.
>>46996747
Don't you dare, anon... D:

>>46996903
>Taking that month-long vacation.
Actually?

>>46996356
Would love botan, okayu and korone this wide aswell

>>46996651
Oh my god, these are so cute!

>>46996957
Yeah, I'm flying out tomorrow. I'll still pop by from time to time. Posting FubuMio at the beach or something.

>>46997040
Goodbye, friend.

Are monstergirls allowed?

>>46997166
Just don't show erect nipples or cameltoes

>>46995758
Dear Seanigger tranny janny:
Just because you're a low-T faggot doesn't mean VTUBERS on the VTUBER BOARD are off-topic. It wasn't even NSFW. Fuck you and your entire bloodline. Pray you never meet me in a waffle house parking lot. I really wanted those Mios

>>46996903
Great voyage to you fubumio enjoyer

>>46997166
As long it's chuuba related

>>46997339
Sana horse pussy got me acting a fool

>>46989604
This with tall hags (Botan, Lui, Mio) and/or Sana

welp I've done enough spicy stuff today. nothing but sfw stuff for the rest of the night I swear

>>46997339
This one turned out a little weird.

>>46997222
https://litter.catbox.moe/xvdre3.7z

Chill and just dl.

Also get 4chanx it'll save you some trouble

Please bully me Koro-San.
>>46995856
You quadruple nigger what the fuck is wrong with you? Kill yourself fucking faggot.

>>46996903
Enjoy your vacation

>>46997536
I got the zip, I just hate seeing fully clothed chuubas deleted, even if it was architecture. There's much worse out there the jannies leave alone.

>>46997494
I'll probably do more tomorrow, but for now I got places to be. See ya vtai.

>>46997639
whoa wtf did I miss that got you that heated?
genuinely questioning

>>46997731
The tranny fuck who is still camping this thread deleting posts that don't break any rules.
Fucking faggots coming off facebook to moderate 4chan.

>>46997801
oh. well I learned my lesson last time talking about them so I'll just leave it be now. not looking for another 3 day vacation

>>46997731
Janny that hates architecture and hags but leaves troll posts and blatent NSFW alone. Always seems to happen around this time of day too. We're all sick of his shit

>>46993053
This is by far the best one I've gotten so far. Didn't even ask for the coloring on the dress.

>>46998148
Box pls

>>46997536
>spoiler
never knew that this extension existed. now I just gotta figure out what the fuck it does

>>46998180
Here you go.
https://files.catbox.moe/5gf2pj.png
Have another Tenshi too.

>>46989604
Pomu in a state of all-encompassing despair.

>>46998148
That looks wonderful anon
now make her ass 4 times the size of her torso

>Stupid fucking jannie deletes SFW architecture
>Fun thread grinds to a halt
Every time. Fuck this guy. See y'all later I guess

>>46989604
chad face pippa like this

her wish came true

>>46998466
I'm slowly learning how to prompt and get goodd results which is nice.
I'm in the minority here of not enjoying wide.

>>46998913
Good on ya, Kanata has always been a staple for me thanks to Coco
I'm playing both side for thin and wide is kino in their own way

>>46989604
Masked torturer Suichan

>Anons getting banned for making ON TOPIC POSTS
HOLY FUCK.
Merge anon is banned again apparently because the post he made about the Bayesian merge was OFF-TOPIC.
FUCK RAMADAN

>>46997536
Thanks a ton for the re-up, I was finally here at the right time to nab it. It's just a shame there's a clear bias in what gets deleted and what doesn't. is there a drawback to making litterboxes last for longer? I've never used it but it seems like you can set how long the link is valid for, why not have it last a day or two (iirc the max is 3 days?)? Especially for unique uploads/stuff that isn't just responding to "catbox pls"

>>46999165
Kanata is cute. I generally do just enjoy small designs as a whole too.
That's also valid. I know I adore sundresses so I will use them often because I can.

>>47000082
>Merge anon is banned again apparently because the post he made about the Bayesian merge was OFF-TOPIC.
Lmao please tell me you're just speculating

>>47000082
I don't understand what the thread is still here for if we aren't even allowed to use it. Just fucking move the thread already so we can actually post.

>>47000189
Based
[Spoiler]Based[/spoiler]

>>47000537
anon...

>>47000537
Point and laugh

>>47000606
Please be patient, i'm mentally ill

I’ve been letting my Kanata with larger assets cook while I’ve been painting, getting to the point do you guys want a zip of them when I’m done? It’s not nsfw she is clothed btw

>>47000341
You know for a fact that I can't answer that in the affirmative. because of glowies
But the fact that a fully on topic post with no image attached and just catbox links was deleted should tell you everything.

>>47000728
I would like the zip.

>>47000728
Gladly, drop the zip

>>47000742
>Can't answer that in the affirmative
I more meant that I can't confirm that.

>>47000728
Drop the link to the zip inside of a pastebin. Apparently linking to catbox isn't enough to stop your post from being deleted.

>>47000846
>>47000907
Ok let me finish painting and I’ll move back to my pc room and zip it up for you guys

>>47000110
Would love for litter to last up to 24h atleast

Here is the original post about the bayesian merge because fuck Ramadan


paste.bingner.com/paste/84quz

Colab webUI broken?

>>47000341
With the below posts anon is basically saying that he has some way to contact merge anon and learned it from them
>>47000742
>>47000909

>>47000994
do I really need to? it's not nsfw at all

>>47002140
Jann hates architecture. Shouldn't be that big of a problem with link only

>>47002140
Merge anon got banned for linking to catboxes in an otherwise fully on topic post.

>>47002327
This post in the bin link got merge anon banned
>>47001111

>>47002503
Literally 1984

>>47000846
>>47000907
https://pastebin.com/HetbEkrf
please let me know if I linked that correctly

>>47002896
You did correct. Thanks for Knight-cup tenshi

>>47003146
cool. now to find a new chuba who's outfit comes out well

>>46989604
I know it's a lot but could we get Ina wearing clothes inspired by the Chinese Zodiac signs?
I'm obviously not asking for everything 'cause that'd be a lot, but something snake related, ox related, rabbit related etc could be cute

Thank you!

>>47003088
Kyo mo suicopath!

https://litter.catbox.moe/dyz0fb.png
Kek

Honestly I think most of the active contributors just got banned. have we considered creating a bunker thread on /bbwai/? Or a different board?

>>47003554
Kek, what was the tags to cause this

>>47003088

>>47003624
They're either banned or they think it's not worth posting any outputs in the thread because it's impossible to tell what content is and isn't safe to post. I'm in the latter camp.
At this point I'd rather just move the main thread somewhere else alltogether. I don't know where but if this keeps up we're just going to die a slow death as less and less people participate in it.
t.okayuposter

>>47003624
The architects have always been the most active posters, that's why the thread was 16 hours long yesterday when they were banned. Several architects including myself post nonfat content as well so that's why seemingly "normal" posters disappear as well.

>>47003554
https://litter.catbox.moe/d0viin.png
>>47003691
Put it in PNG info to see the monstruosities

kizuna AI

>>47004323
I would rather die, but we can always have a mirror thread and just crosspost if needed

>>47003088
Did you use Cardosanime for this?

whats the reasonable amount to use for highres steps mine is at 75

>>47004845
???
I use 15..

>>47004689
No, this is a personal mix which is mostly Anime25D. It seems CarDos Anime has a much stronger contrast.

>>47004845
Most use half of their sampler steps

>>47004845
>75 hi-res steps
jesus fuck mate. I run 30 regular steps and then 15 hi-res.

bbw model anon and users, specific issues to the fats model completely aside for a moment. You lot are right. You're very, very, very right.

trying to grug Marine's new outfit in hll4p3ep8 for about twenty minutes. Her fucking pirate hat, ascot, skirt, etc. are all GLUED to her even worse than hll1, which I actually complained about...

>>47005104
Switch to a similar enough hll3.1 merge. tada.wav...

hllanon, I don't know if it was the Lion thing, some Unet fuckup, or what, but something isn't right.

>>47005104
anything under 25 looks like shit and blurry as hell to me I just kept going up to see if it actually got better

>>47005243
You've got two of my prompts there to look at and I'm not particularly good.

It must be something else. 75 hi-res steps is ridiculous unless you're using DDIM and after diminishing returns. The DPM samplers practically stop doing shit completely after 30 steps. Euler a, pointless to exceed 40 steps, unless you want to squeeze an extra bit out to 50, but it's Euler a, so it will start doing weird shit. Using the same amount of hi-res steps as initial steps is the "regular" overkill, and going ham with like 100 total DDIM steps is the mega overkill. 75 hi-res steps with a DPM sampler? Insanity.

TL;DR Check your negatives
PS. Latent will give you blurry output at lower denoise strengths unless you go all the way back to Christmas time commits...

>>46996651
かわいい!!!!

>>47005243
You have another issue, its not the high res steps.

I dont know what gets banned, fuck it.

>>47005472
>The DPM samplers practically stop doing shit completely after 30 steps
That's actually only true of 2M. 2S a karras and SDE don't converge.

>>47005472
I mean I can settle for 30

>>46996651
>https://files.catbox.moe/sqrkd8.png
damn brat, disappearing out of existence like that !#

>>47005243
>anything under 25 looks like shit and blurry as hell to me
my lolis are fine

>>47004323
If it comes down to it I'm still in favor of the other place, especially since hdg already has a presence there

>>47005591
AH, gomen, I didn't realize that, thank you. I use 2M most often, must've filed it away in the noggin wrong.
>>47004323
y u heff to be sad?

>>47005104
>bbw model anon and users, specific issues to the fats model completely aside for a moment. You lot are right. You're very, very, very right.
>trying to grug Marine's new outfit in hll4p3ep8 for about twenty minutes. Her fucking pirate hat, ascot, skirt, etc. are all GLUED to her even worse than hll1, which I actually complained about...
This is where I get to say I told you so!

>>47005784
>the other place
/bbwai/?

>>47006001
No I'm not even sure what that is lol
The place in in the gitgud under other sites

>>47005955
I'll take it like a champ. It's the first time I actually try prompting something base up with one of these instead of dropping an old schizo prompt into it to see what it does. That seems to be where it shines...picrel a new schizoprompt I wrote from scratch...I couldn't get anything nearly as goofy looking with hll3.1.

>>46995188
This one look amazing

>>46993053
I tried mixing all three
Child Watame colouring while drunk aunty Tenma is soaking her feet in the water. Did 50 gens while I made food.
https://files.catbox.moe/rsafkj.png
https://files.catbox.moe/0c72cn.png
https://files.catbox.moe/zq2gq8.png
https://files.catbox.moe/z3nmu8.png
https://files.catbox.moe/15cumv.png
https://files.catbox.moe/15cumv.png

>>47007197
kek, thanks.

>>47007197
Damn it double posted the last one
https://files.catbox.moe/xji7lk.png

>>47005955
By "I'll take it like a champ" I mean yes, I was wrong and I'm not above it, kek. Feel free to rub it in my face.

https://litter.catbox.moe/jpd4z7.png
Anyways, now that I have the image I want. The {{{biggest}}} colab update is incoming in a few minutes. I might've broken something, there are actually too fucking many options for me to test now...

>a bunch of the setup happens asynchronously, more on that below
>you NO LONGER NEED THE SHORTCUT TO /stable_diffusion/ in your google drive. you can keep it if you want, but it does nothing. it's all handled via cloudflare and huggingface
>there's a "Reset Settings" buttons that loads whatever settings/styles.csv I have on huggingface
>bunch more music nobody will probably listen to but me while I debug it, kek
>the option to disallow Google Drive access actually works now, but your outputs will all only exist on the Colab! also it obviously can't load your previous settings nor your LoRA folder without Drive access!!!

it installs all the pip requirements, and downloads the repositories that webui usually clones, while your selected model is actually downloading. pretty sure Google has these in a datacenter with a big fat pipe and SSDs, might as well beat up on them!!!

the biggest and only remaining slowdown is actually...extensions, and only on first launch. while I was testing, I had every single extension enabled - it took 98s just to "Load Scripts" during the initial startup, which must've been installing requirements.txt for one of them that just had a laundry list of stupid shit it needed. I'm going to figure out which one it is later and make it optional...the only one I'm sure it ISN'T is the civitai one, because I rewrote it myself and know that it has no requirements.txt, kek

shit nobody cares about but me:
>bing chilling error handler and a bunch of other silly HTML moved out of the notebook because it can load it externally just as fast. links in the prompt if you're curious, functional /vt/winamp in there too. also bing chilling is now a BSOD. kek. I'm going to make the other regions fun too on lunch breaks during the week or something
>winamp got an upgrade. you can fucking put milkdrop visualizers in a bloody imaginary console window while you wait for it download your anime girls, we truly live in the weirdest timeline.
>uh...if you do whip the llama's ass, your Google Drive mounts at the very beginning. it's not measurably faster, but it's considerably less annoying.

anyways, because of how fuckin slow SD first launch is, the fact that a bunch of shit might still be downloading asynchronously is a non-issue. your checkpoint will always be done downloading before sd ever gets to loading it...

hopefully vtai can recover at some point

>>47005862
>y u heff to be sad?
Because I've seen this movie before with /become/ and /ag/ and I know how it ends.

>>47005104
>bbw model anon and users, specific issues to the fats model completely aside for a moment. You lot are right. You're very, very, very right.
Yeah, something is indeed wrong with it. Although anons have gotten p2 merges to work relatively well, so it's possible that epoch 8 was overtrained

>>47008043
>Although anons have gotten p2 merges to work relatively well
NTA but thank you for this info, I will have to give mixing another shot.

>>47004979
So your personal mix was also used for these?
>>46993283
>>46996230
>>46996747
My card is too weak to actually train or mix models, I've never managed to wrestle Anime25D(by itself) to get that the halfway painterly look you did though,so I was wondering if it's doable through promps.

>>47006186
See that's the problem instead of prompting like a normal person you are inputting so many token into the clip encoder that it actually just explodes

>>47008299
mixing models is done in RAM

>>47008299
Model merging only requires graphics power if you are doing an auto merge like Bayesian or Auto MBW which need to generate images in order to determine the objective value of said merge

>>47008357
>have 8 gb VRAM
What's your recipe?

>>47008322
that is me trying to prompt like a normal person anon, kek
>>47008299
model merging is entirely I/O and RAM based. every single model this thread downloads from grugger? merged on a fucking Ivy Bridge CPU. there's no way your rig is shittier than mine. none. if you have 16GB system RAM and aren't using a hard drive then you have everything you need to merge a model.

>>47008299
Yes, these were done by me. You can change your output to be more painterly using Loras (check style loras on Civit) and in limited amount through prompting, but as the others have already said, it might be easier to just mix your own personal merge or use one of the many mixes that have already been uploaded. The real limitation when merging is your RAM, not VRAM.

>>47005174
It's just bad balancing - it had too many pictures of Marine in the dataset. I should have just deleted half, instead of relying on multipliers. NAI already knows her, anyway.
I'll fix Marine, Mori and the issue with wings next time.

Just in case the doomposting is actually warranted this time, I just wanna say, it's been a wild journey watching /vtai/ go from using NovelAI and trying to replicate it on local hardware to where we are now. No matter what happens, stay awesome and keep prompting, /vtai/.

>>47009503
Thank you king.

anyways, mom's spaghetti. please loudly yell racist epithets into the void if I broke something. otherwise, it should be about twice as fast to launch now if you aren't actually installing any extensions.

>>47009842
Yeah it's been a hell of a ride regardless of what happens. These past few months with you all has been some of the most enjoyable I've had in awhile.

>>47007920
I'm still watching this movie with /nasfaqg/ and our messiah has risen from the dead, on easter, of all days. glass half full okayuposter.

>>47003457
I don't think my model knows how to do those prints on dresses, even at weight 1.9. These are the sensible ones doubt I'll get something with mouse, dog, monkey_print
https://files.catbox.moe/yjop3h.png
https://files.catbox.moe/knlax3.png
https://files.catbox.moe/nssw1t.png
https://files.catbox.moe/drdi42.png

I was with my family, is this really the end?

>>47010938
Yes
/vtai/ is over
For the 13th time

>>47010809
>monkey_print
A FUCKING MONKEY
>>47010938
no, okayuposter is being a doomposter

>>47010938
3-4 active posters were banned.
according to posts on /vtai/
One of them was banned for making a fully on topic post that only linked to catboxes that had {{{bigger}}} art to show off the capabilities of a new model merge

>>47008526
If you still mean me, I combined Anime25D with a little bit of Counterfeit and Silicon. It's still mostly Anime25D. If you prefer e.g. CarDos, you can probably replace Silicon with CarDos, or do another merge. Counterfeit was mainly used for concepts.

>>47010938
and it's your fault, retard

>>47011056
Seems like it might be for real this time, a month ago we averaging three or four threads per day.
Last week, every day onvr2smly had two threads.

>>47011547
oh yeah, it WAS his fault
>>47011553
quality > quantity?

>doomposting with 53 IPs
in 15 some odd years on and off at this dump, you know what's killed more fucking generals than anything else? people who whine about the end

>>47011547
>>47011758
Yes, but because it's my fault I refuse to let /vtai/ die.

>>47010938
No, it's not over. It's just Sleepy Sunday with Easter holidays and a couple of temp bans added on top of it.
I'll still be posting here even if I am the only left doing so. Might just end up being a FubuMio thread tho, kek

>>47012106
I'll be here schizoprompting center openings or posting silly shit

kek, I remember when /vtai/ was just the 2 mios, okayu poster, towaposter, me and the requesters

granted there were other anons that came like once a week or so.

btw, I don't know if the anyhentai anon is arrounf I merged and pruned the new version with hll3.1 and 4 is uploading to HF

question can you make folders on HF or you need to do that on local and then commit it?

the power of tummy will prevail.

>>47012106
I'm new but I'm planning on sticking around and trying to prompt a bit more as I get used to how this stuff works.
It's good practice for my final goal too.

>>47012203
Original fat Mio poster here. I'm sure Smug is around here too. I remember when people thought we were the same person. I miss fashion Watame poster :(

>>47012203
as in can you make folders using the web interface? same workaround you'd use on github. try to make a new file. like /newfoldername/.gitignore or something - you can make it a blank file. it will make the folder by force to stick the file in it. once you have a model in there you can delete it.

So if I wanted to start doin this myself what kind of specs do I need? Also how much storage will all these programs and stuff take? I'm on an old laptop from 2018 so I don't think I'll be able to do it.

>>47012499
Does your laptop have an nVIDIA GPU?

im newb, how do i make fauna without loras?

>>47012769
Carefully.

>>47012488
Oh, will try that, never have to use it like that before but with this I don't want to move my local stuff around, thanks.

>>47012402
A lot of people came and went

>>47012769
>VTuber Prompt List: https://rentry.org/vtaiprompts
read the OP my guy, and have a vtuber model

>>47012769
>download any hll model
>best quality, ceres fauna, hololive, uuuuuuuuuuu
that's it

>>47012499
>>47012769
both of you:
>>/vtai/ Resources
>https://rentry.org/vtai/ | FAQ: https://rentry.org/vtaiFAQ | VTuber Model Mixes: https://rentry.org/vtairecipes | Embed/Thread Archive: https://rentry.org/vtaiarchive | VTuber Prompt List: https://rentry.org/vtaiprompts

>>47012891
lmk if u get stuck with anything else with huggingface's weird ass git implementation

>>47012203
>towaposter
man I haven't posted towa in so long...

>>46989604
I need more of the phase girls as slaves you don't understand its like a drug PLEASE

I think I (eggs)'d too hard. if nobody finds a Colab bug within the next half hour I'm goin to sleep, kek

>>47013267
I don't know how to call you half of the time because other anons know you for different names

>>47012639
No. AMD Radeon
>>47012963
Thanks.

>>47013441
>AMD
it's over, he doesn't know...

>>47013505
I actually went out and got a nvidia card after exclusively using amd cards for over 10 years just to proompt.

>>47013505
I'm guessing I'm out of luck then.

>>47013441
Well, you can always use the collab

>>46999661
>ask the ai for jason
>get phantom of the opera
kek
>>47013426
I was being nostalgic...
>>47013441
depending on the Radeon - it may not be worth it. if it has 8 GB dedicated VRAM and is Vega or newer, you can try either the DirectML fork on Windows eh. iffy or if you're comfortable with Linux, do it that way because it actually works more like it does for GeForce users
otherwise:
>>Colab for Complete Retards
>https://colab.research.google.com/drive/1STL60qfoY-iSlhRb9zFETRLTqhNbznRf

>>46999661
dozo. also FUCK everybody that included COVID images in LAONININO whatever SD was trained on, fuck everybody that drew anime girls in fucking COVID masks. fuck COVID. I hate them I hate them I hate them I hate them

>>47012499
I started with a 1050TI less than 4GB of VRAM. Took 7mins to load up SD and can only make up to 720x720. Made coomer shit for myself just fine.

Took less than a day to setup and figure out how to prompt.

Reserve about 15gb for models and Loras I guess

goomba..

>>47013818
I've never tried to prompt characters who own masks with masks on, is it really that bad?

>>47014008
Looks like she needs braces to me anon....

>>46989604
Short stack Gawr Gura.

>>47013639
>>47013626
>>47013846
I'll look into all of this. Thanks lads.

>>47014045
yes. you can't say "mask" without getting a fucking surgical mask if they don't already have some mask associated with them. and even then, you'll still get a surgical mask. the mouthveil lora works well enough though.
>>46998292
oh nyo

>>47014062
she's a shark anon

>>47014269
That one needs a helmet

>>47014269
>Interrupt

>>47014229
{{{sadder}}}

Is there a way to consistently avoid fusing when using latent couple or regional prompter?

>>47014045
It's frustrating that I have to run a lora for any kind of mask that isn't a surgical mask. It's frustrating since I'd presume people who want a mask want an actual mask and not those face diapers.

>>47014401
at leas shes not from springfield anymore?

>>47014269
...is it weird that I kind of like it?

>>47014499
Top left looks pretty good. Surprisingly, I like the weird traces of pink in top right.

>>47014499
negative prompt 1girl, 3girls, 4girls,..

>>47013639
>webUI spit outs a bunch of errors and the site doesn't work at all

>>47014499
Use Multidiffusion one.This one does it in a way that avoids fusions on multi-lora prompts. Downside hard to get character interactions (at least I haven't figured it out yet).
https://github.com/pkuliyi2015/multidiffusion-upscaler-for-automatic1111.git

not really a doompost but I've been posting a lot less this year
not sure if it's just memory bias but I've found that prompting random/funny stuff like corn or mushrooms has decreased a lot

>>46976356
Really nice, thanks anon

>>47014453
ogey
>>47014499
controlnet, openpose + posex
>>47014706
let me get my crystal ball for you...

>>46985491
Really nice, thanks anon

>>46989773
>Even AI is having trouble with hands.

>>46989604
Bae cosplaying as Chun Li or Cammy.

>>47014706
wait, are you a new user? or are you someone who's been using it trying to let me know something is wrong with it? or a troll
either way, RTFM first, and second
>bunch of errors
yes, I know what all of those are. copy and paste the traceback into a text file or something

>>47013426
catbox please

>>47014748
Corn apocalypses and things were fun because we were still figuring out just how far you could push NAI into shitpost territory because chuuba posting on its own was limited. With the advent of chuuba specific models and good looking mixes and loras, it became far more possible to post whatever chuuba you wanted so people got more into normal prompting since more people could prompt their oshis. The newer models are also significantly harder to pull goofy shit out of than NAI was so it became less fun to try. At least in my opinion.

>>47003457
Clothing inspired by the mighty hog

>>47015122
That would explain why I can't make physical anchors anymore.

>>47014706
>Felllow amd user
I decided fuck I want to ai art one day so after a google search I found this guide, and used it as a last ditch effort since I was tired of trying to wrangle pythons all day.

https://medium.com/@fanis.spr/fast-and-easy-way-to-use-stable-diffusion-on-windows-nvidia-and-amd-bcb728af29db

And guess fucking what, it worked.

You'll probably get some memory errors the first time, so you'll need to follow that optimization guide, and you'll probably still need to figure out how to use collab since youre basicly stuck at 512x512 with out it since amd is shit for this, but you can at least gen some fun shit.

>>47015122
>>47014748
plus I was a big chunk of the sillyposting and now I coomerpost too much
>>47015204
sampler. you can make anchors with euler a a lot easier than the others for some reason, but okayuposter still makes DPM anchors too

>>47014942
dozo

Good morning vtai
>>46982939
>>46988720
>Both of you fell asleep immediately from exhaustion after sex last night
>You woke up this morning to this sight
>"Papa, mama, why are you two not wearing clothes when sleeping?"

>>47015376
>you can embed wavs in rats now

Maybe I'm a slowpoke,and I'm missing out on not using collab, but thats fucking amazing.

>>47015376
Baeutiful.
think you can give me another one with her lifting her little skirt while showing a disgusted expression?

>>47015437
my wife is also a lolicon

>>46993119
You going to post this image in every thread?
>>47014942
dozo2
>>47015568
sure, but I just put cammy in the oven

>>47014269
>>47014008
Every single gen you post is the most uncanny bizarre shit, yet I can't help but feel endeared to some of these shorks

>unprompted center opening
>all the way down
in what universe does cammy look like this?
https://media.tenor.com/AVeWTGaaNPcAAAAC/ron-swanson-parks-and-recreation.gif

>>47015682
>You going to post this image in every thread?
Signs point to yes. And yet somehow all these "reference" posts aren't off topic while actual vtubers are.

>>47015045
Hope is not too messy, I was trying to get a clear bra under her shirt
https://files.catbox.moe/2x92sa.png

>>47015568
yeah, switching back to chun-li, I'll get ya yer disgusted upskirt. the second an unprompted jacket shows up I'm fuckin done with the idea lol.

>>47015007
https://pastebin.com/WbVKmzpm
I have no idea what its happening, it opens the login screen but nothing there works, even trying to use a wrong user/password does nothing

Space Moom
>>47015122
I do miss a bit of the silly posts >we would get back when NAI was a thing,
>The newer models are also significantly harder to pull goofy shit out of than NAI
That's the unfortunate fact that with models getting better people will shift towards them. It's definitely still possible to do silly stuff. But it it is definitely harder now.
>>47015437
Good morning Luluposter.

>>47015929
Think you can prompt a Mai cosplay without too much effort?

>>47016005
It also doesn't help that trainers are obviously making coomer models instead of more broad/general models

Wider image kinda works lol.

>>47014786
My laptop doesn't have enough power for controlnet lol
>>47014732
Thanks, this stuff looks useful.

>>47015980
Anon those errors are normal, it's because it tries to pull up a bunch of pages/assets before you've actually logged in.

Now...since you didn't sanitize that, kek, go ahead and stop the currently running "let's fucking go" cell and start it again, since everyone knows that ngrok URL. oh, you already closed it. good. anyways, it happens every time and means nothing. you should be able to log in just fine with whatever your username/password are.

also uh. fuck I forgot. oh, I remembered

1. firefox complains about ngrok.
2. shitty anti-virus programs like, at least avast, will complain about it

>>46989604
This anon in the drawfag thread has a great idea >>47014655
I'd like to see this with any lush chuubas, Mori and Gooba in particular, onegai

>>47016145
We also used to bully and ignore coomer requesters instead of encourage and enable them and allowed them to take over. When's the last time you even saw a goofy request to do?

>>47016273
ah, natsukashi...

>>47016057
Mai Shiranui from KoF? Baelz or someone else? did her with Flare before, I can find the prompt and just plop Baelz into it

>>47015568
I'll roll 1 or 2 more times but upskirts are so much easier with controlnet and I don't have it running at the moment...

>>47014269
HEY YOU GUYS

There's an old reference

>>47016273
We were coping, coomer shit was hard

>>47016378
Bae cosplaying as Mai Shiranui is what I meant.
>did her with Flare before
do you still have those images around your HDD? think you can catbox them?

>>47016005
I can't believe AstroMoom is a phoneposter...
>>47016145
>making coomer models
There's are still some broad models around, granted some of them are built with more specific themes instead of full on general use. But I think what really contributed to the nai era having a good mix of coomer and silly was tthat with nai being the primary model most people used back then allowed >us to really break it down and push it to the limits.

>>47015568
lucky you
>>47016461
I catboxed them when I originally posted them
https://files.catbox.moe/kwkrv4.png
pretty sure you'll find more if you chase catbox_kwkrv4.png down in palanq or warosu

>>47015437
Do we lie to the Twolus or respect them enough to tell the truth...
Flip a coin, Luluman

>>47016492
Pick up that can!

>>47016461

>>47016461
ah dynamic cfg my beloved, you and your worse than usual hands...

>shiranui flare bleeding through because mai SHIRANUI
CLIP, I hate you

>>47016799
>feathers on the combine soldier.
Moombine...

>>47016273
I've been doing the creative prompts

>>47015376
>>47015682
>>47015929
>>47016378
>>47016548
>>47016916
>>47016992
Thanks for these anon, sorry if I was a bit overkill with these requests.
>>47017092
kek.

>>46965406
was out all evening yesterday and haven't gotten on until just now, but holy shit this helps so much. I'll see what I can do about baked-in standing stances, but I'm now able to do cool shit with Yuri.
also damn son that second image you linked with the bra/panties stuff was hot as hell

>>47015980
did it work?


TL Note: I haven't played KoF since before I could drive...so since before half of (You) were born. Uh, all I remember is red kimono looking thing and booba. Good enough.
>>47017227
I always bite on fighting game requests. especially if it didn't come with a picrel like we don't already know who chun li is or something, kek
o7

g'night hos

>>47015821

Anyhentai models up
https://huggingface.co/pozoleai/chuubas/tree/main

Will add them to the rentry
https://files.catbox.moe/b6ktsd.png

Haven't played with it a lot but it has most of the popular models in it