/vt/ - Virtual Youtubers

Please don't straight up just fucking kill me, miss Shiranui.

Pavolia Geyne

So far (30girls) is amalgamating 30 different holos into 10
except for mori who is clearly too powerful
https://files.catbox.moe/vbgtm4.png

>>43702793
Kronii is fairly distinct too.
And... Mare Bellofiore? Why's he got tits?

Big God Mion

https://files.catbox.moe/asg4cf.png
https://files.catbox.moe/j65xri.png
https://files.catbox.moe/jsfb0p.png
https://files.catbox.moe/lvkaca.png
https://files.catbox.moe/n2b184.png

>>43702559
https://litter.catbox.moe/3v68kk.png
w̵̧̛̛̪͈͍̺̘̖̞̟̦̣͇͍̜̠͉͠ͅḭ̬̠̱̤̪͞t̸̶͙̭͇̫̻͎̬̺̭̱̪̱̮̕ͅh̡̝̳͙̭̮̻h̶̫̯̮̦͓̪̯ó͇͖̦̘̘̱͟l̸̵̥̥͓̙̙͍̝͔̳̦̭̹̬d̠̟̟̮̥͝ ̸͎͉̦͎͢ͅa̠̱̟͖̰̖̙̖̘̕͘͡s̴̨̠̺͙͔̼͟͞p̸̨̗̺̮̯͚̜̩̮̬ę̴͚̗̗̻̻̣̫́́ç̺̳͓̯̗̜̙͍͟͡t̷̥̖͚̱̝̠͉̣̻̦͈̺̥͈̭̹͢͡ş̶͙̠͉̖͈̘͜͞ ̡̡̺̘̣͚͈̜͓̰̝̤͓̝͎͈̝̙̹͟h̴̢̢͖̺̜̬͙̫͇̪̯͇͚̜̳͚̱̪̙́ȩ̵̸̧̤̜̠͉͇͈̩̮͝ļ̵̟̼͍͔i̸̸̷̛̻̮̤͚͘o̷̙̯̝̭͉͖͎̳̘͖̤͎͉̩͜ͅs̴͚͉̰͕̥̝͇̕

The discovery of fire

Yay, new bread

Anchor doko? ANCHOR DOKO???

>>43702932
No requests in this thread.

>>43702932
Moribikebros...

>>43702902
You're playing with metaphorical fire right now anon.

>>43702932
You're the anchor now.
I'll have some Sana tummy please.

>>43702628
YOU HAVE TO EAT ALL THE EGGS

Request anchor

>>43702886
NOW I'M GOATIVATED

Deploy GarlicxyGoatsGaming

https://files.catbox.moe/etuah1.png
https://files.catbox.moe/y7on6f.png
https://files.catbox.moe/7rgpzb.png
https://files.catbox.moe/jdgjsj.png
https://files.catbox.moe/1f82cu.png
https://files.catbox.moe/91xbpo.png
https://files.catbox.moe/sjf2b1.png
https://files.catbox.moe/eao2f9.png
https://files.catbox.moe/b4aogx.png
https://files.catbox.moe/tg0wge.png

>>43702886
Can't wait for {{{{{{{{{{BIGGER}}}}}}}}}} God Mion

Goodnight /vtai/

>>43703010
Fuck
I'll try not to forget my image this time

>>43702932
Ina as picrel pls ;_;

Good evening /vtai/

I'm gonna try training a lora tonight. Which one of these installers do I use? https://github.com/derrian-distro/LoRA_Easy_Training_Scripts/releases/tag/installers-v4
>install_kohya_ss_ui_v4.bat
>install_sd_scripts_v4.bat

No worries, anon. I'm making more bread for your oshi. It won't make her {{{bigger}}}. Probably. You trust me, right?

>>43703166
>>43703032
SHIT I got sniped! >A<;;

>>43698836
>>43698973
>>43703166
I wanted to go sleep but then I saw your request but it didnt turn out well and I'm tired so ima go sleep slightly mad

Good night vtai

Maybe we don't deserve an anchor.

(replying again incase it was missed)
>>43702273
Not really, batch size doesn't correlate to speed. You're basically telling it to update the model less frequently, which can help smooth out or "average" the training, because at smaller batch sizes the output could be more noisy. If anything it will potentially take longer. To maximize speed your goal would be determining the smallest batch size that can still train cleanly the fastest. Of course that's easier said than done because you still have to worry about LR, scheduler, the dataset itself, etc.
Someone feel free to correct me if this is inaccurate, I am (as I'm sure many others are) self-learning a lot of this ever since the SD leaks, lol.

>>43702932
No anchor today, anchorfree day

Happy Rabbit.

Requesting heavily pregnant pippa surrounded by fbi agents

>>43702932
No requests allowed in this thread!

>>43703211
>install_sd_scripts_v4.bat
This one
w̴̵̡̢̗̬̳̙̹͔͍̦̣̻̳̭̯̘̯͡r͖͈̗̦͓͎̘̻̳̰͢ờ̖̟̠͇̞̫̖̮̠͚̬͔̰̳͚ͅn̡̤̪̘̣͍̙̜̜͓͓̣̠̣̠̹̘͘͞͡͡ǵ̴̨̬̣͓͍̘̘̦̮͈ͅ ͉̬̠͙̬̟͚͉̪̭͈͍̖̀͟ͅͅa̴̴͉̮͓͕w̶̺̞̦̳͕͡ͅͅņ͓̪̟͕̺̠̣́ͅs̮͈͓̩͕͔̪̕e͚̦̮͖͜͠r̶͖̹̥̦̗̦͕͙̘̬̙̗̹͔ͅͅ ̹̰̟͕͙̥͓͙̬̘͇̤̜͕̣̀͘͘̕͜ḩ̷͇̯̭͉͎̺̘͚̹̭͓̠̘̪͡͝i̢̭̠̘̙͔̞͈͙̱̫̪̣̦̹̗̰͢͠͡͠v̵͎̦̭̮͖͚̳͞ȩ͈͍͎͈̪̭̭̗̝̮̮͍͈̦͜͠m̸̵̢̻͓͇̯̙͍̮̘̣͙͔̘͇̫͕̤̭͟i̶̵̘̥̯̯͙͜n̴̡͔̺̜̫͘͠͡d̯̤̼̺͉̱̹̟͞͡

>>43703619
requesting someone to do the request anchor

>>43700245
Thank you anon. Greatly appreciate it.

>>43702907

>>43703032
kek

>>43703426
Probably made her too big, yeah?
But this is it from me today. I gotta sleep.

>>43703316
Thanks for the info. I guess if you want to maximize this smoothing effect beyond what fits into the VRAM, you'd use gradient accumulation? Although I googled a bit on the optimal batch size and apparently huge batch sizes will lead to detrimental results?

>>43703128
I see you got the variable skin colors going too. Very cute kusogaki goot. May your prompts be motivated! Have you considered posting some of your better gens to pixiv?

>>43703131
>>43702886
{Bigger} God Mion (I can try cranking it up to architect level later)
https://files.catbox.moe/qr5d59.png
https://files.catbox.moe/xw0nrx.png
https://files.catbox.moe/ios7lf.png
https://files.catbox.moe/mbifah.png
https://files.catbox.moe/ra5j33.png
https://files.catbox.moe/nkj1gz.png
https://files.catbox.moe/5p772b.png

>>43704243
Huge batch sizes would do that I assume, because then you're averaging out too much. Gradient accumulation I believe is only to save you some VRAM.

>>43704316
>(I can try cranking it up to architect level later)
Yes please.

>>43703128
and the NSFW garlics
https://files.catbox.moe/atub65.png
https://files.catbox.moe/y4dx85.png
https://files.catbox.moe/xgb2he.png
https://files.catbox.moe/jta9ll.png
https://files.catbox.moe/dyuzgs.png
https://files.catbox.moe/efkn6y.png
https://files.catbox.moe/fq5fi8.png

>>43703888
You're welcome, it was good pose training, had to bin a lot of the early ones so got some good practice

>>43704316
Mmmmmm, big mio, yummy.
I've considered hosting it somewhere, if nothing else then just for my own reference and organization, but I've got no experience with it so probably not something I'm likely to do anytime soon. I've also never used pixiv so dunno if that's the right place for it or whatever.

Fuck that last sentence had to be rewritten so many times, it's getting too late
Imma join the rest of the eurofags and sign off for tonight, may you all have a blessed thread.

Goodnight /vtai/, see you all tomorrow bright and early (so catch me back here in like 8 hours)

>>43703692
>>install_sd_scripts_v4.bat
>This one
Thanks takomind...? That's the one i installed at least.

>>43704561
Towa... <3

Good morning
https://files.catbox.moe/gc2o90.png
https://files.catbox.moe/rwmew4.png

>>43703032
Reimu and Ina's backs and booty.

>>43704641
>https://files.catbox.moe/efkn6y.png
Nice. I love tail masturbation.

>>43703010
Sana tummy

>>43703032
Can I have an Ina with small breasts, and with the "she's my best friend since elementary school and now that we're adults she has a big crush on me but I'm so dumbass to notice it" trope?

>>43705065
Thank you, that is a nice tummy.

>>43704048

>>43704641
Goodnight, anon. Also, Yammers :3

>>43704938
Good morning, Lulu poster.

>>43704697
No problem
w̷̯̝̝̤̞͈̝͙̘ͅà̷̡̞̖̫̙̺͚̳̫͈̫̞ŗ̢̬̟̫̠̖̩͚̬̫̺̪̼̻́ṛ̷̛̤̙͚̺͈͟͠á̸̴̙̞̭̳̥ǹ̻̳̱͕̤̝͚t̵̢̧͍̜̦̼̤̺͘͝ͅ ̴̨̼̯̮̙͘͞ḁ̤̗̟̭̙̀͢c̶͔͉̰͙̀͟͞͝c̪͍̟̗͔̹̣̖͓̤̭͘̕͜ͅu̶̗͙̣̦̜̰̜̥̯͓͓͚͎̟̙̫͜͡m̷͎͚̟̖̥͇̗̜̟̖̘̪͕͈͈̻͜ͅù͈͍̳̫͈͈̺̩̳̞̫̬̩̭͍̜̦́ͅļ̤̟̮̹̘̩̣͚̱̭̰͘͟ͅạ̵̴̩̭̘̯͔͈͍̥͈̰͈̯͔͓̻͡t̵̨̫͚͓̱̦̗̺̹̀e̢̛̹̯̯̣̩͖̖̗̦̯̮̙̗̗̱͟ ̡̗̰͓͙͍̀͡h̴͍̜̝̣̫̜̼̮̝͔͙̜̥̀͘a̵̧̤̣̝͓̩̱͇̝͔͖͎̘̳̠̱̖̙r̸͓̮̫͈̭̬̟̞͚͙̭̮̟͉̰͕͓̜̹ŗ̵̴̞͇͖̜͔̟̫̼̳͡o̵̯͓̺͍̠̠̪̗̝̥̼̭̠͝ͅw̵̵̧̮̭̪̜̰̫͇̙̼͕͟

>>43705065
can we have more?

>>43705375
Very Chinese.
Who is this supposed to be? Blue Fauna?

>>43705440
Oh, I thought this is green? My color picker says green hue at least.
And yes, it's Fauna.

>>43702932
Bratty Loli Reine.

educate yourself (but wholesome because feesh wants you to succeed.)

>>43705375
NTA, but it's a quite bluish shade of green. Fauna's green is very yellow.

>>43705375
Wow, this looks beautiful. Love the artstyle.

>>43702932
Crusader Marine praying
Re-requesting Marine in a Napoleonic era military uniform (French preferred)

I'm a paranoid retard afraid of torrenting, is there any way to use a website to do it for free or is the only free way to grab one of the leaked models?

>>43705869
Is this edited or did the AI really make that text?

>>43705930
Yes. The AI wrote kek beneath breaststroke.

>>43705930
Obviously edited.

>>43702932
Can I have a Mumei riding into battle atop a giant owl or gryphon? Preferably wielding a lance.

>>43705922
>I'm a retard.
Colab for Complete Retards: https://colab.research.google.com/drive/1STL60qfoY-iSlhRb9zFETRLTqhNbznRf

>>43705922
You dont have to grab any torrent, a lot of the stuff you want are avalible for free on huggingface.
Are these prompts too much?
w̴̢̹̗̗͕̹̝̖̮̼͙̦͞o̷͉͔͙̹͚̞̖̙̮̥̻͘͞n̨̝̭̬̘͕̥̠͖̰͖̦͈̙̱̟̮͜͞ḏ̸̸̳̯̺͓̬̹͜͠ę̝̠̣͎̬͎̻̤̟̭͎͇͔͜͝͝r̸̷̹̺͕͓̰̪̬͍͠ ̵̝͍̞̻̻̙̖̦̤͍͙̘̳͙̩̀á͈̱̥̕͜͜͝r̶̻̯̱̦̤͎̦͍͙͇̰̀̀͢͠ͅt̵̸̩͙̥͎̻͚̘̯̺̖͙̣̘͎̻̞͎͎ͅį̷̝̯͚͙͚̦͖c͈͓̬̞̭̠͔̮̺̤̀́͡ͅų̴̻̤͈͙̗̼̩̖͜l̨̨̪̣̠̥̰̰̮͠a̶̛͖͕͉͔̳̳͎̠̙̹̞̪͉̣̣͟t̵̹̘̖̮̞̕͠͡ḛ̷̷̗̙͔͍̬̻͕̮̫̠̰͎̤̟ ̷̵̡̻̠̰̺͓͕͎̖̰͉͈͕͍͟͡ͅh̫̞͖̞̥̠͍͕͔̠̺̯̯̝͎̪̱̕͞͠͞ͅi̴̷͇̺̺͔̬̮̝͚̪̠̺̮̕͟ͅm̴̨̛̹͖̫̳̦͎͡ś̥͔̣̮̣͎̠̣̱̦̜̦̱͙͠ę̡̯͚͕͔̟͈̦̜̭͎̳̞̥͉̭̙̀́͟l̳̣̭͍̰͎̝͙͖̱̭̤̳͖̘̦̕͝͞͡f̘͓̖̹̪̜͍̦͖͙̩̰͓͎̥̤͜

>>43706108
I'm enjoying them quite a bit.

I'm really liking this Lora
https://files.catbox.moe/f6ghq1.png

>>43706108
Looking thru your prompt. Did you manually put brackets around each token? Is there even a difference between just grouping them up in one big bracketed group vs doing this separately?

>>43703254
Bless you sleepynon
Inja is so perfectly horny

>>43706175
I just kept adding adjectives
>>43706267
Who knows. Control and force of habit.

>>43706108
I always appreciate eldritch horrors

if there like a recommended base negative prompt thats least likely to fuck with my images? And amy recommended negative embeddings? I had been using bad prompt v2 but was browing civitai and saw people running like three others in the same prompts

learning in the library with kanatan

>>43706108
>spoiler
Nah they look really cool

>>43704641
Spicy garlic! Sleep snug goat anon. Pixiv's easy to use and has dedicated tagging for AI art, so the anti-SD faggotry is kept to a minimum. Up to you, but I think it would be cool to see more stuff from this board on there. Compared to the other AI boards on the chans, at least, I feel like this one produces serious quality.

>>43704591
>>43704316
{{{Bigger}}} God Mion
https://files.catbox.moe/ken3fr.png
https://files.catbox.moe/j23mud.png
https://files.catbox.moe/1ybawk.png
https://files.catbox.moe/7jnoi7.png
https://files.catbox.moe/6v750z.png
https://files.catbox.moe/xzmisj.png
https://files.catbox.moe/vt0ji9.png

>>43706877
catbox please

Gura browsing the local candy store

>>43706879
Thanks

>>43707046
they got fruit mentos?

>>43707012
https://files.catbox.moe/j7pf7h.png

>>43702932
>>43703032
Squid Kid Ina or Octillery Ina

>>43706954
Turbo sex, bellies are the best.

>>43707116
that's a sensible chuckle from me, anon

>>43707409
got carried away with the logo and finding beeg transparent mentos, forgot that I was supposed to prompt new landscape mio panic. kek. oh well, tomorrow is a new day.

>>43705375
catbox please?

>>43707541
Here: https://litter.catbox.moe/vredt1.png

>>43705384
Im busy rn but i'll drop some from time to time

My wife, so pretty

>>43707853
Is she OK?

>>43707782
hng

>>43707940
Just some food poisoning, she'll be fine

>>43707853
Is your wife the melty flying one or the tiny slug in the foreground?

So I've got my sd scripts installed and read through the LoRA Training Guide a few times. I think I get it but I'm not sure.

So I want to make a lora for Okayu's sololive outfit. I already have a folder specifically for that getup so I don't think I need to go through the gathering/sorting a dataset steps.

Now do I go through my folder and tag the images one-by-one with the wd14 tagger, then prune the taglist sof Okayu's traits?
Then after that (and after I set up the image directory) I see there's three training settings from various anon's, which one do you guys recommend?

>>43707853
>>43708258
Nightmare fuel and I mean that Ina good way.

Where can I get the 4x-AnimeSharp upscaler from? I haven't been able to find it from here and haven't seen much of it from outside this site

>>43702902
And she used it to burn her pubes

>>43702932
Oh come on this don't even link to previous thread

>>43707116
Yeah they do

>>43708508
Look in the upscale rentry, theres a link to the wiki with the upscaler

>>43708508
https://upscale.wiki/wiki/Model_Database
ctrl + f

>>43708258
More
>>43708331
>Now do I go through my folder and tag the images one-by-one with the wd14 tagger, then prune the taglist sof Okayu's traits?
Yes your data needs captions.
>Then after that (and after I set up the image directory) I see there's three training settings from various anon's, which one do you guys recommend?
I used my own personal experience with embeds and hypernetworks and chatgpt and picked my own settings. Its mostly defaults though. But I havnt saved a config of mine yet.

>>43707046
What the hell the lightning is so good

>>43708655
Looks like our baker was new, drunk or both.

>>43708655
OP is a baka

Oops all tentacles

>>43703032
Vtubers if they were artillery

>>43705296

>>43708734
Okay thanks.
Is there a faster way to tag than copying the tags from tagger, pasting them in a notepad then rinsing and repeating for every image?

>>43707782
thank you, that's great tummy

>>43709041
I'm pretty sure there should be a tag by batch option or something

>>43708705
>>43708696
tyty

>>43707621
reminds me of clannad. love the aesthetic.

>>43707382
>>43706954
Even {{{Bigger}}} Mion
https://files.catbox.moe/cvp2bz.png
https://files.catbox.moe/mgevgw.png
https://files.catbox.moe/qjiwfl.png (I think smug architect Mio poster would be proud of this one)
https://files.catbox.moe/mui96i.png
https://files.catbox.moe/t1a1dl.png
https://files.catbox.moe/v1vcue.png

>>43708876
>>43708258
>>43707853
AI is perfect for Lovecraftian stuff. All the little inconsistencies and flaws common to AI work just become Non-Euclidian details beyond comprehension when you're looking at something eldritch.

>>43709041
If you use grabber to DL images from a booru you can config it to scrape the tags into .txt files alongside your dataset.
If you dont have them tagged you can use the WD1.4 tagger extension inside webui to generate you .txt files. Its accurate enough.
Then DL the dataset manager to easy mass prune tags from all the .txt files.
See >>43682149 for more. Im going to bed. oyasumi

>>43709041
Check out the dataset tag editor extension

>>43709189
Based bigger poster.

>>43709417
Link: https://github.com/toshiaki1729/stable-diffusion-webui-dataset-tag-editor

>>43709138
There is a "batch from directory" option with the wd14 tagger but I dunno what setting to use with it to actually get what I want.
>>43709376
>>43709417
Thanks, I'll check thos out. oyasumi!

>>43709342
Morbidly tempted to do Fauna Shub Niggurath, if I had the time.

>>43709671
Like...a sexy Shub Niggurath?

>>43709749
Is there any other?

>>43708705
>>43708696
Alright im retarded, how do I install it?

>>43709942
Put them in stable-diffusion-webui\models\ESRGAN

Garouma LoRA gives me Misadventures of Flapjack vibes

>>43710168
>horateerss

Are you taking your pills, anon?

>>43702932
gigantic tits watame

Sexually frustrated Space Lulu
https://files.catbox.moe/yk3p1e.png
https://files.catbox.moe/qy5qo0.png
https://files.catbox.moe/ud00e4.png
https://files.catbox.moe/2kvoa0.png
https://files.catbox.moe/l0pv1q.png

>>43702902
This actually looks pretty good.

>>43710248
backhand sugoi

>>43708668
kek. doooooo doooooo doooooo doo doo dooo wahhhhhhhhhhhhh

>>43710637
I hate that I can hear this post

Is it just me or does controlnet canny make some pretty derpy faces I have to inpaint after?
>>43700953
>The Coomerlicious vTubers
What is that model?

>>43709584
>Use the batch tagger for wd14
>Check "Sort by alphabetical" and "Escape brackets"
>Select "wd14-vit-v2-git" under interrogator
>Link your dataset up top, hit interrogate. This is by far the easiest and fastest way to tag datasets, especially when you get big ones
>Wait for it to finish tagging all the images in the directory for you
>Use either the Web ui data set editor or BooruDatasetTagManager. I use the latter so I don't know how good the other is.
>Review all the tags. It's not perfect so add some if you see it missed it, remove ones that are plain wrong as well as ones to do with your character

>>43702628
Can I has some Lize Helesta? Has there ever been Lize Helesta? Bonus points for fully nude.

>anons who know a thing or two about tagging or training are afoot
oh shit. I have a bunch of .txt already with the tags from the boorus. do any of the auto-tagger things allow me to add onto those tags, rather than replace them outright?

>>43705711
"I TOLD YOU I WANT HAMBURGERS AND PIZZA FOR MY BREAKFAST, IF YOU DON'T BRING THEM I'LL TELL MY PARENTS TO FIRE YOU!"

>>43704938
Look how soft lulu booty looks. I wanna squeeze and bite it.

>>43711301
Correction needed

>>43703032
I just want some Amelia ass pics to make a teamate day even better.

>>43711301
Wow her budding breasts are the greatest delicacy

>>43711188
Thank you. I did something like that and then used the dataset tag editor to prune character tags.

Now I'm at the part where I can actually institute the training but the rentry is frustratingly vague at the very end. I don't understand how many concept folders I should have for my dataset of 35 images. And even after that, the guide doesn't say how to actually start trainin the lora, it just lists some settings and that's it.

>>43710248
Can you do wasteland scavenger Gura in this style? Hoodie on, mask and backpack

>>43702628
Question: My drive isn't far from getting full, could I transfer stable diffusion to another one without any problems?

>>43710248
Where's her big fat tail?

>>43708954

>>43710248
No, and you CAN'T make me...

>>43711950
Cute

>>43711650
>I don't understand how many concept folders I should have for my dataset of 35 images
If you're just training one thing you do one folder. You're just training her single outfit, so it'd be something like
>[Sololive] ----> [Repeat#]_WhateverYouWantToNameThis ----> (your dataset here)
Everything else depends on the which method you're using to train it, so I'd need more info to help you with that.

>>43706877
Where is the private time on the corner of library.
>>43710391
Has she tried anal perhaps?

>>43711950
Damn that’s cute

>>43709749
The problem is that you and I probably have very different ideas of what makes Shubby sexy.

>>43702149
Nice one. Ina only needs to be a fair bit above the water nude for it to be like she is talking to another begin.

Please, help yourselves to some more cute Rosemi feet art:
https://files.catbox.moe/deqsy5.png
https://files.catbox.moe/1dbkk1.png
https://files.catbox.moe/wb4oqh.png
https://files.catbox.moe/sbdkwb.png
https://files.catbox.moe/qw2iwo.png

Really hard to get good results with hires fix enabled, for any chuuba for whom I'm trying to make art; picrel is only good one I could get so far for Wosemi. I'm also still struggling with art for Nene Amano, even on default 512x768 with no hires, please be patient with me

>>43702932

Not an art request but

This model 8528 got eradicated from hugginface before I could grab it

https://huggingface.co/852wa/8528-diffusion

Had really pretty gens, especially for backgrounds


Does anybody have a copy of it I can steal?

>>43711993
Ok thanks. So I set it up like
>[Sololive] ---> [5_conceptA] ---> (my dataset)
As for the second party
>Everything else depends on the which method you're using to train it, so I'd need more info to help you with that.
I don't really know what method to use to train it. I installed sd scripts earlier for what it's worth.

>>43706927

sorry for my lack of posts today anon's was rather sleepy all day still fucking am sadly but monster and coffee will help me survive

>>43712268

why is your post

spaced like

that?

>>43712268

>>43712268 (me)
Here have a suisei as repayment for my indescretions

(It's sfw but over 4mb)
https://litter.catbox.moe/l1l8mi.png

>>43712458
Fuck off I'm literally on my phone and driving

>>43710309
Sheep tits
https://catbox moe/c/9o7iep

>>43711496
Yessir

>>43712587
PUT IT DOWN DUMBASS

>>43712587
>anon crashes and dies asking for a model that he can't even install until he gets home
Many such cases!

>>43711340
Me too anon, just look at those!
>>43712000
Not yet
>>43711351
>>43711605
Corrected
"How bout you eat this instead?"
https://files.catbox.moe/1g1l13.png

>>43702932
Male vox with female breasts

>>43712699
lovely sheep tits my friend

>>43702932
TimeRyS embodied as a literal fusion of Kronii and IRyS.

>>43713349
Built for tentacles

>>43709189
>>43709571

Last batch before I get back to schedule
https://files.catbox.moe/gr0pec.png
https://files.catbox.moe/yjzh5n.png
https://files.catbox.moe/otiyiu.png
https://files.catbox.moe/0rfg98.png

>>43712403
Yup. I usually just name the "conceptA" part the same as the preceding folder since I don't make complicated loras. Also, if you're just training the outfit and don't need to have it learn Okayu (the images you've posted already have her, so you've got that covered) don't bother removing the tags that have to do with Okayu herself, like hair color etc. If you just remove the tags to do with the outfit, it will learn the outfit and you can even use it with other characters later so they can wear her outfit with little conflict.
>I don't really know what method to use to train it.
Looking over the training guide again, it's been changed from when I installed mine when it used windows powershell, so I probably can't help you with this part as well as I'd hoped. I'll go looking through it anyway, but it'd probably be faster just to read up on it yourself instead of waiting for me. Sorry about that one.

>>43706954
>>43709189
a body like this is built for constant sex

>>43710248
>Gura is dropping several things
This image is far more accurate to its portrayed chuuba than anything else I have seen in these threads.

>>43713645
{Fitter} God Mion
https://files.catbox.moe/69per0.png
https://files.catbox.moe/edto4l.png
https://files.catbox.moe/8femv1.png
https://files.catbox.moe/71a6ue.png
https://files.catbox.moe/hv1z79.png
https://files.catbox.moe/p60snb.png
https://files.catbox.moe/ecwyju.png

>>43702932
Dommy dictator Fauna

>>43703032
>>43702932
Nurse Gura giving the viewer a checkup.

>>43702886
>>43704316
>>43706954
>>43709189
>>43713645
AND HERE'S TO YOU, MRS ROBINSON

>>43712755
Thansk

>>43711993
nta, I thought you could bin the images this way? same concept name, different # of repeats. no?
>>43711297 onegai
>>43711743
use a colab that doesn't rely on your drive to store github repos and models that can be downloaded in a few minutes. there's one in the OP you're replying to...

>>43713729
Gotcha, thanks.
>Also, if you're just training the outfit and don't need to have it learn Okayu (the images you've posted already have her, so you've got that covered) don't bother removing the tags that have to do with Okayu herself, like hair color etc. If you just remove the tags to do with the outfit, it will learn the outfit and you can even use it with other characters later so they can wear her outfit with little conflict.
Hmm... so if I understand this correctly, pruning Okayu-related tags won't do much. But pruning outfit related tags will let the outfit be used on other characters too? I'm not really sure who I might put in the outfit other than Korone.
>Looking over the training guide again, it's been changed from when I installed mine when it used windows powershell, so I probably can't help you with this part as well as I'd hoped. I'll go looking through it anyway, but it'd probably be faster just to read up on it yourself instead of waiting for me
Will do. I'll check it over again.
> Sorry about that one.
You have nothing to apologize for, I greatly appreciate you taking time out of our day to help me with this.

>>43713645
Thank you for these and all your other bigger posts.

>>43714172
I have a player in my PF2e game who has made it his goal to be stabbed by a beautiful prostitute in every location the party visits. God bless you for giving me the arabian stabbing whore.

>>43703032
Passionate sex with Rosemi-sama

I also would like a ahegao expression

Please

>>43713235
Could I get this but Fauna? And an even fatter ass?

>>43714607
sure but It'll be like 10 minutes curranty fucking around with controlnet to make elira dominate a guy

>set everything up
>try to make ina
>It keeps making her cunny despite me not mentioning anything remotely close to it
Bros... I think the AI has a taste...

>>43714790
On the contrary my AI usually gives Ina a figure comparable to Suisei's.
Try putting "loli" in the negatives or change model if you don't want cunny.

>>43714790
Skill issue

>>43714521
I guess I'll be nagging the shit out of you soon if I can't figure things out on my own end. I should probably mass rename my grabber tags txt's, have wd/deepdanbooru take a go, and then just use a really dumb FOR EACH loop to concat them all into one .txt file

I need to find more regularization images that aren't Marine...and my "oh shit, I actually made some proper center opening with the AI" folder is now at 66 images. ignoring the 200+ I grabbed off boorus. I know people just throw shit into the trainer and get good results, but I want this thing to "just werk" with the one activation word and additionally to respond to breast strap, covered collarbone, criss-cross...I forgot, and bare shoulders.

>>43711203
hll3 is pretty good at her
https://files.catbox.moe/j6cdqh.png

>>43715080
>Suka Blyat on bottom left
Fucking kek

>>43712414

>>43714285

>>43714790
skill issue
>>43715149
why do you think I had to make sure my proompt PNG -> JPG script could handle Unicode properly? kek the prompt box is for shitposting.

>>43715080
>I should probably mass rename my grabber tags txt's, have wd/deepdanbooru take a go, and then just use a really dumb FOR EACH loop to concat them all into one .txt file
I like your funny words magic man.
Are you having trouble mass tagging or something?

>>43715205
Oh my. Now this one is absolutely perfect. You have my utmost gratitude

>>43702902
Aside from lack of beauty mark and wrong amount of crown diamonds, I genuinely would've thought this wasn't AI, it looks really good, there's just always that feeling of dissonance and confusion and "why is this character illustrated super well in this style and in this situation" that these always tend to have.

oh and Ina's third foot

>>43715438
No, although because I have to use colab for it, I'm not going to mass tag until I'm all done. I have a dataset that contains images that came with tags from boorus - I went over those manually while I did things like shoop out patreon watermarks - and there was quite a bit missing. I want to combine the mass tagger's output with what was already on there so I just have a huge set per picture to prune, rather than having to scrutinize the image and see if anything is missing.
and I'll definitely need the mass tagger to tag the AI generated ones. plenty of shit that I DIDN'T prompt ends up in them.

Ok maybe I'm retarded (very likely) but seriously, what the fuck is this guide?
https://rentry.org/lora_train#training-settings

You go from "How to set up the image directory" to "Training settings" and that's it. There's no information on how to actually use all the steps they outlined in the guide to make a lora. Maybe I'll go head over to /h/ and complain so they can call me a retard too.

>>43715582
>Ina has a beauty mark
Wow I never noticed

>>43715765
What more do you need?

>>43714532
You're quite welcome

>>43714172
{{{Fitter}}} God Mion
https://files.catbox.moe/oyp0j1.png
https://files.catbox.moe/b34pje.png
https://files.catbox.moe/n7jrm5.png
https://files.catbox.moe/12l1cx.png
https://files.catbox.moe/uxelfb.png
https://files.catbox.moe/5l5rfs.png
https://files.catbox.moe/0vnlmk.png

>>43715853
How to actually institute the training of the lora? The guide teaches you how to make a dataset but there's nothing about how to start the scripts/training.

>>43715878
Hot damn that is snu-snu material.

>>43715946
If you installed the sd-scripts like it said, it tells you how to put the training settings in.

>>43716063
I installe sd scripts hours ago. There's like a dozen different scripts in the folder.

>>43711203
>Can I has some Lize Helesta?
Sure but I'm currently in the process of working with tentacles so pardon for the mess.
https://files.catbox.moe/sih8m9.png

>>43702932
>>43703032
MILF Marine and/or Marine with (you)r many daughters

Marine fusion w/ Ina or Gura

Is there any syntax difference between + and , as delimitators, or are they identical?

>>43716300
run_popup.bat

>>43711979
Goodnight, /vtai/

>>43716336
use commas
>delimitators
huh. it didn't underline in red. I always called them delimiters.

>>43711496

>>43715537
have some more

>>43714500
>same concept name, different # of repeats. no?
You can do that, but that's to have a lower number of repeats on different images. Say you have 50 good images, and 10 you think are okay as fodder, but you don't want the lora to pick up too much from in case it lowers the quality. Then you can have TWO folders in the [Sololive] folder, and give the lesser quality images their own folder with a lower number of repeats, giving them less sway in the finished lora but allowing it to learn some very basic things from them.
Example.
>[Sololive] ----> [10]_WhateverYouWantToNameThis ----> (your GOOD dataset here)
>[Sololive] ----> [4]_WhateverYouWantToNameThis ----> (your LESS GOOD dataset here)
That would apply both to the same lora but run less of a risk of ruining your final result with sketchy quality images. Both being in the Sololive folder but splitting your data set.
>>43711297
Not to my knowledge, but the w14-vit-v2 is usually so accurate you shouldn't have to worry too much about that, I believe it's based off danbooru scans to begin with. You don't actually have to do that much pruning of mistake tags, and the tags it does miss usually have very little effect on the lora. You COULD save the txts you already have, separate them from their respective images so they don't get overwritten by the tagger, have the auto tagger generate new tags for those images and then cross reference your old txts with the new ones with a tag manager to insure you're not missing anything though.
>>43714521
>pruning Okayu-related tags won't do much. But pruning outfit related tags will let the outfit be used on other characters too?
Correct. What you prune is what the lora focuses on learning, that's why it's important to have things you don't want tagged in the image so it doesn't learn them on accident. Even so, if the learning rate's too high, you have too many repeats, etc, it may end up learning them anyway. It's very finicky. An example of mine would be when once I had a "phone" tagged, but the lora latched onto it and would try to gen the character holding a phone, or something square like it unless I pu it in the negative prompts. If you solely focus on the outfit, it has a much greater potential to work not only better on other characters, but should also reduce the number of images you need for it to learn it. The more complicated the subject of a lora you train the more images you need to train it.
>>43715765
>You go from "How to set up the image directory" to "Training settings" and that's it
I swear it used to have more. It used to call for powershell to train, looking at the dates on the bottom it's been updated recently. Continuing my reading and troubleshooting on it now. Mine doesn't run popup.bat correctly, so I'm looking into that. If you can get that to work for you and have more questions though let me know.

alright faggots im not in my car anymore

does anyone have the 8528 diffusion model

>>43716396
Goodnight FubuMio anon

>>43716679
>>>43714500(You)#
>>same concept name, different # of repeats. no?
ty, but wouldn't be "Sololive" in this case, I'm NTA
>>[Sololive] ----> [10]_WhateverYouWantToNameThis ----> (your GOOD dataset here)
>>[Sololive] ----> [4]_WhateverYouWantToNameThis ----> (your LESS GOOD dataset here)
So those two folders can have the same name, oui?
>>43716679
>Not to my knowledge...though.
Plan currently is to let the autotagger do it, and then concat it with my previous txt's. Banking on any good tag manager should be able to clean dupes out

sir, please understand, I want to make a LoRA that will take ANY outfit - and unzip it clean down the middle, neck to navel. kek

>>43716651
Happily. I wonder what she's planning with an expression like that. You have my gratitude.

i was tried to dig through my collection for another random cute image but its all porn

>>43716679
>Correct. What you prune is what the lora focuses on learning, that's why it's important to have things you don't want tagged in the image so it doesn't learn them on accident. Even so, if the learning rate's too high, you have too many repeats, etc, it may end up learning them anyway. It's very finicky. An example of mine would be when once I had a "phone" tagged, but the lora latched onto it and would try to gen the character holding a phone, or something square like it unless I pu it in the negative prompts. If you solely focus on the outfit, it has a much greater potential to work not only better on other characters, but should also reduce the number of images you need for it to learn it. The more complicated the subject of a lora you train the more images you need to train it.
Okay, I'm gonna prune things again using this advice and take out all the outfit related tags. This is a pretty small dataset (34 images total), so should I take out other tags such as background and expression related tags? Also I do notice that 6 images have a swing so maybe I should take that out too to avoid your phone example.
>swear it used to have more. It used to call for powershell to train, looking at the dates on the bottom it's been updated recently. Continuing my reading and troubleshooting on it now. Mine doesn't run popup.bat correctly, so I'm looking into that. If you can get that to work for you and have more questions though let me know.
I'll try the popup.bat after I prune things again and I'll let you know how that goes!

>>43716396
Night!
Im out too. See ya vtai

>>43713177
further correction needed thanks

>>43717481
Unable to comply, subject ran away

>>43717570
*trips and falls*

>>43714790
ina?
cunny?

anon what in the fuck are you prompting in there shes built like shes 19
>>43717570
>>43717655
catch her while shes down

>>43716859
>wouldn't be "Sololive" in this case
I know, just continuing my example for ease of understanding and to avoid confusing anyone by jumping around with names.
>those two folders can have the same name, oui?
The "whatever" folders absolutely can and probably should, yes.
>I want to make a LoRA that will take ANY outfit - and unzip it clean down the middle, neck to navel.
That should be a fairly simple one assuming you can get a nice, high quality dataset for it. It tends to have more trouble and need more examples and training time the more fine details it has to learn.
>>43717198
>should I take out other tags such as background and expression related tags?
No. Absolutely do not do that.
Think of the lora training like this. The trainer can detect what's in the image based on the tags you provide it, it finds what it knows cross referenced with the tags, and tries to exclude what you've told it is already in the image. The tags (though in this case since you remove them, the "visual stimulus") you DO NOT provide it in the tags, it will attempt to "learn" as whatever you tell it they are. In this case, whatever it cannot identify in the image would become the "sololive outfit". If you remove Okayu's tags, her facial expression, and her clothes, it will learn all of those things as the lora "Okayu sololive outfit" and attempt to apply those whenever you add it to the prompt instead of just her clothes, her face, hair, AND clothes. Everything you remove BECOMES the "outfit".
If you were to try to train a character for example, had 100 images of her smiling and removed the "smile" tag, it would assume
>This character is always smiling
and even if you prompted something like "frown", if the lora was trained hard enough for long enough it would have trouble making her do anything but smile. For most loras you're aiming to train it hard enough it can reliably get the details of what you want right when prompted, but not so hard it starts to refuse to change anything about it. Like asking it to change the color of the outfit. Maximum accuracy yet remaining flexible.

>>43717864
I'm the first half but that second half has helped me a lot. I'm gonna want to make sure "navel, stomach, collarbone, center opening" are all removed from my /perfect/ images, and control for the weaker ones by remembering to add in things like "covered collarbone" or "partially undressed". Sucks that there's no opposite to "bare shoulders"...

>>43717864
I see! Thank you so much, that's a really clear and concise exlanation. I think I get it now.

>>43714607
I am so so so sorry this took so long to get to you

Selen seggs

Idk what she's holding but I really like how this turned out

>>43718197
That's okay. It's also pure sex!

>>43718254
Numbing lubricant

>>43718291
here is another then tried to make the ass {{{bigger}}} for you

>>43716302
Is that the Rindou art style?
>>43715115
Thanks!

>>43713177
(realism:1.1)

anon?

>>43718443
WOAH MAMA, catbox please?

>>43718197
what the fuck this actually looks really good

Good morning, /vtai/.
Had a batch of robospider Kroniis baking while I was sleeping. And they all look... kinda homogenous? Like I'm scrolling through gens and they all look mostly the same. Not sure how to properly explain it, but it's like there's nothing truly outstanding. Body types don't vary that much, spider parts look all similar. I don't fucking know. I'm not very satisfied with this batch, compared to preceeding FBK and Anya gens.

>>43718527
sure thing thing https://files.catbox.moe/ohkn2l.png
btw do you mind if I try and add kronii into this?

>>43718607

>>43718631
These ass prompts are simply Magnificent Anon

Please keep going

>>43713177
Wow inverted nipples, you prompter is very cultured

>>43718086
>>43718049
No problem. One more detail I want to add just in case, since a lot of the AI threads are getting this fucking horribly wrong somehow. You should try to train your lora to get best result around a weight of 1. There are a lot of people training them to the point you have to weaken them to a .6 strength, but even then that usually makes it hard to customize the image and severely burns it out as well. The sweet spot is getting whatever results you're most satisfied with around the .8 to 1 weightings so it remains nice and flexible while not burning your image.
If you don't understand any part of that right now that's fine, don't worry about it and just focus on training. You'll learn what I mean when you start getting the finished products out the other end.

>>43717570
A little cum dripping down would be perfect

>>43717655
>>43717777
>>43718506
It's on everything I prompted, including the Lulus

>>43706039

>>43716599
Thanks anon

>>43718932
Ohh box?

>>43718898
now she gets an ice cream

>>43718932
Whoa, that's pretty fucking cool.

>>43718857
Nah I understand wholeheartedly, thank you for reminding me.
t. browsed civitai and saw lora that seriously mentioned "you have to use it at 0.3"
motherfucker you overfit the hell out of it

>>43718898
is fine, i didnt have your model and did not carefully read the prompts
"eh creamsicle will probably work"
pizza delivery for anon, sign here please

>>43718979
https://files.catbox.moe/fpen9r.png
ControlNet does most of the work

>>43718757
holy sex

good morning spider anon


i would request with panties off but i know what your gacha rates are

>>43718681
Thank you. The only thing better than a giant ass is two giant asses, I'd say do it.

longcat

I really wanted to make her deformedly long but AI was not cooperating

>>43719099
>i would request with panties off but i know what your gacha rates are
I actually got one that didn't have panties on completely unprompted. Gimme a minute to upscale it.

>>43719256
kronii spider prompt with 5* gacha rating and 30% critical sex rate

>>43719099
>>43719324
Enjoy
https://files.catbox.moe/b6z4z7.png

>>43718994
>>43719422

im kinda surprised it didnt spider her hair

>>43718631

>>43718757
>>43718994
>>43719256
>>43719422
Arachronoids
>>43718784
It's all gacha
>>43718885
That would be perfect, but the AI gave me waterfall most of the times
>>43718990
"Here, please forgive uncle okay. Don't tell your parents about what I just did"

>>43719454

Sex rat

>>43719499
catbox?

>>43719422
>She's pointing to it
holy sexo
I'm imagining her gesturing and saying stuff like "come on don't be shy. What, afraid a single spider?" and when you enter she locks her spider legs around you
sorry for blogpost

>>43719557
and then she fucks you to death and eats you

>>43719437
That would have been neat. Anything to make this batch less generic.

>>43719550
https://litter.catbox.moe/12dg22.png

>>43719499

Yeah I'm going back to generating Fubuki. Kronii has been somewhat disappointing and I still can't put my finger on exactly why.

>>43719613
was going to try and do a little prompting myself to this

on the one hand, thank you for not giving your model fusions useless names

on the other hand what the hell is counter cream pastel

>>43719789
It's... counterfeit + creamypesto + pastelmix?

So the run_popup.bat works for me, which is nice I guess.
It just asks so many questions and I dunno what to say to 80% of them ¯\_(ツ)_/¯

I'll just go through what it says for fun. Some of you might find the struggles of stoneagechama entertaining.
>Do you want to load a JSON config file?
EZ
>Select your base model
Uh... I usually use HoloCreamsafe but lets go with HLLCounterfeit nuhard-fusion-pruned.
>Select your image folder
wat. "sololive" I guess.
>Select your output folder
...
"sololive" again?!
>Do you want to save a json of your configuration?
Fuck no.
>Which optimizer do you want? The original optimizer was 8bit_adam, or admw for AMD. Lion is a new one that was recently introduced
Lion?!
>Are you training on a SD2 based model?
MAYBE?!
>Are you training on an realistic model?
No...?
>Do you watn tot use regularization images?
nani the fuck
>Do you want to continue from an earlier version?
What earlier version?!
>Do you want to flip all your iamges? It is supposed to reduce biases within your dataset but it can also ruin learning an asymmetrical element.
No because this outfit has a lot of asymmetrical aspects to its design.
>Do you want to change the name of output checkpoints?
... No?
>Do you want to eset a comment that gets put into the metadata?
No comment.
>Do you want to train only one of unet and text encoder?
NANI THE FUCK
>Do you want to save a txt file that contains a list of all tags that you have used in your training data?
Sure I guess.
>Do you want to use caption dropouts?
N A N I T H E F U C K
>Do you want to use noise offset? Noise offset seems to allow for SD to better generate darker or lighter images using this than normal
Sure! Why not?

TL;DR: Here are the results
>lora_train_popup.py: error: ambiguous optoin: --lr_scheduler=cosine_with_restarts could match --lr_shceduler_num_cycles, --lr_scheduler_power

>>43719137
I'm sorry for some reason it's just not generating. which fucking confuses me cause I was able to gen stuff like this yesterday but come today bam fucking nothing

>>43720466
Strange. Well either way your stuff is great no matter how many chuubas you put in.

>>43720208
I'd say just stick to training with the base NAI model, but that might be just another superstition. And your model most likely isn't a SD2 model, I think. At least NAI isn't. Not sure what the error message is about.

>>43720513
catbox?

>>43720593
https://files.catbox.moe/x80ejt.png
I had to scribble in the phoenix

>>43720208
update sd-scripts

:)

>>43720466
holy sex
catbox pretty please?

>>43720725
daww

>>43720725
Cute!

>>43720725
Now done with tongue, anon °<°;;

I have spent an unhealthy amount of time on this sole image
https://files.catbox.moe/xjispq.png

>>43720731
sure thing https://files.catbox.moe/ge8esq.PNG

>>43720916
Thank you!

>>43720810
Too young for tongue stuffs :)

>>43720565
Thanks, I'll train on ye old NAI.
>>43720670
huh?

>>43720208
Welp, Okayu-anon it didn't turn out as easy as i thought it was. I assume all that wall of text is related to you making a lora? And not a model.

>>43719518
Indeed.
https://litter.catbox.moe/znhhg7.png

>>43720810

Good morning!

>>43720996
My uohs go unheeded... T∆T

>>43702932
Any chuuba as a thicc robot in latex

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

>>43721051
It's probably easy, I'm just retarded. I'm trying to train a lora for the poisonya syndrome outfit.

>>43720544
wow yeah its only that pose that won't work it seems

>>43721183
me on the right

>>43721082
FELLATE THE GUN, YOU PEAFOWL WHORE >:U

Good night.

>>43715115
Edit over here:
>>>/e/2760015

:(

>>43720208
>>43721044
It's me, here to annoy you with information again since I just got mine working.
>>Select your base model
Usually you select a lower model in the chain for this to insure good compatibility. Many select NAI since a lot of models are built and mixed off of it. This thread probably wouldn't mind if you used hll since there's so many model mixes you guys use based around it. That would be very nicely compatible and get you good results for both HoloCreamsafe and HLLCounterfeit since they're both mixed off that model. NAI is the old reliable though, and usually the best option. I'd recommend NAI.
>"sololive" I guess.
That's correct, a lot of people make the mistake to select the #_Concept folder, but you want the one right before that.
>"sololive" again?!
That also works, you can also set up a separate folder for only outputs if you'd like.
>Are you training on a SD2 based model?
I'm unsure if hll is SD2. You might just want to train it twice, once saying yes and once saying no if you've got the patience for it. That other anon said it isn't though, could just trust him honestly.
>Do you want tot use regularization images?
Don't bother with that. you can skip it.
>>Do you want to continue from an earlier version?
If you've trained your lora and found it's underbaked, you can say "yes" here and select the lora it gave you as an output to continue training instead of starting over. It saves you a LOT of time and headache, though depending on how under baked it is you might want to change the # in the folder name you use to train. An under baked lora can be put back in the oven, an over baked lora is burnt for good unless you use a older epoch (more or less something like a "game save" made mid way in the lora's training).
>Do you want to train only one of unet and text encoder?
You can ignore that for now as well, it's more complicated than you'd need for your first lora.
>Do you want to use caption dropouts?
Ignore that as well.
>>lora_train_popup.py: error: ambiguous optoin: --lr_scheduler=cosine_with_restarts could match --lr_shceduler_num_cycles, --lr_scheduler_power
Delete your sd-scripts and LoRA_Easy_Training_Scripts folders, then run the install_sd_scripts_v4.bat again. I had similar problems earlier, that should give you a working version. Most likely your sd-scripts folder is from an older version and isn't meshing with the files installed by the .bat.

>>43721249

A friend asked me why I wasn't using AI to make anime girls eat hamburgers, and you know what, he was right. Have this

>>43720856
It was worth it.

>>43721222
Holy sex. Looking forward to the zip if you do more of these.

>>43721249
blimey

>>43715878
Big {Goot} Mion
https://files.catbox.moe/4gvon0.png
https://files.catbox.moe/xcm7yb.png
https://files.catbox.moe/j9eqoh.png

>>43721336
why did she do it :(

>>43721411
heh you already know btw did I ever share the kanata and rushia one?

>>43721394
Yeah I just copy pasted his error into the search bar (which will work most of the time btw, protip.) and got it from Github
https://github.com/derrian-distro/LoRA_Easy_Training_Scripts/issues/18

>>43721249
there we go

>>43721453
She looks like she's screaming in pain...

>>43721407
Good man.

>>43721485
You posted a few but not a zip.

>>43721553
probably just had an nd into the thigh

Okayu
https://files.catbox.moe/88q7r8.png
https://files.catbox.moe/ai00bg.png
https://files.catbox.moe/2i3es0.png
https://files.catbox.moe/7k0hlh.png
https://files.catbox.moe/1f6x0y.png

>>43721529
Ballin'
Remember: No teeth! :y

>>43721590
https://www.youtube.com/watch?v=p3kJ6SU3ycs

>>43721589
ah ok. well here that is https://files.catbox.moe/r0b3m1.zip

>>43721720
Box pls

>>43721395
https://youtu.be/VziMy4yOzak

>>43721719


require correction

>>43721719
Thanks.

>>43721108
Morning!

>>43721191
Ah, so you are doing exactly that. Great, I wish you best of luck. It's me btw, the anon you talked to before and the tried to replicate poisonya outfit the other day.
https://files.catbox.moe/a2mas8.png

>>43721596
yessssssssssss
>>43721394
Thank you so much! I'll try this again following your instructions.

>>43721485
ok kronii and fauna together are just not working well. need ya'll to suggest who kronii should be replaced by I like fauna more so she gets to stay