/vt/ - Virtual Youtubers

Request Anchor

>>83819294
RE-request >>83723045 Yuu Ayase getting drunk and flashing her puffy nipples

>>83817208
I'm not familiar with Tsuna but I wanted to give her a try, she's cute
>https://postimg.cc/gallery/VdD2Njq

>>83819253
Nice OP! I rather use this one than the Biboo pic since both, the Biboo pic and previous thread's OP were mine. I appreciate the anon suggesting it though. I'll have to post more cute Bibs in the future...

>>83686875
not so much a main poster but going mad fighting animation using comfyUI last few days
progress so far on fauna titty drop
https://files.catbox.moe/8mxks6.gif

>>83819294
Requesting Millie cum hat or similar on any other chuuba. https://files.catbox.moe/au1fcy.png.. Shion has a witch hat...

>>83819294
>>83757265
This is great, could we get some tanned Panko, Shiina or Pippa?

>>83819294
Chuubas trying to fit two dicks in their mouths

>>83819294
Futa (not too big) Gigi and Biboo touching tips

>>83820236
Ohh nice!

>>83819294
Pekomama boobies please

>>83817063

>>83691991
https://files.catbox.moe/71vz5q.png

>>83813672
> nice yuri gens though that first one is really good, worth the time I would say
Yeah I get to get exactly what I wanted by doing photoshop, also building the composition is kinda therapeutic working on it, just like this Mumei, IRyS Fauna gen. Put on a video and do some work.
>>83817757
Thanks, with that it should complete the VShojo list for Pony
>>83820236
Nice to see more people trying out motion, hope you can pave the way for the new tech

Night lads, don't open the door for strange women.

>>83822201
im working now on a v2 for geega so that its less stiff with advice i got

u mas alway go maek promp BEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGGG

https://litter.catbox.moe/n70ak1.png
https://litter.catbox.moe/8cz3mc.png
https://litter.catbox.moe/b0ob30.png

She is very happy for having a leak on her roof kek

>>83819349
Thanks for the bake anon!
> flux or edited
It was just flux, I just inpainted the girl next to duck to be Subaru. Raw gen if you are curious
https://files.catbox.moe/j5zl7c.png

>>83819663
Sexo, as an avid foot lover. thank you.
>>83822201
Damn so many tiddies in one pic, I kneel
>>83822252
G'night, dang there goes my weekend XD
>>83822612
SexoRys love the 2nd one the most

>>83819294
Snek tits

>>83819294
Wife Shiori kiss in bed.

>>83819294
Botan riding a skateboard

>>83823159
Always wondered, do these "exuding carefree vibe", "bold aesthetic", "modern feel" etc. actually do anything in nlp?

>>83822201

>>83702148
https://files.catbox.moe/lxeiud.png
https://files.catbox.moe/bwuk1r.png
https://files.catbox.moe/69mbn9.png

>>83822252
Goodnight
>>83822612
Sexo IRyS, didn't know goo anon got an animagine LoRA too
>>83823159
Nice and cozy with a huge mug of hot chocolate

Goodnight /vtai/

I haven't been posting recently and it's because of these.
Imagine a beach restaurant that has robot maids based on Hololive girls as busty waitresses.
This is the 1st batch. There will be another one after this.
Aki - Now, it's Aki who is the good or bad girl
https://files.catbox.moe/ekd5cy.png
https://files.catbox.moe/mtjpey.png
https://files.catbox.moe/yhng0g.png
https://files.catbox.moe/gkwnxc.png
https://files.catbox.moe/ypqo20.png
https://files.catbox.moe/9r74uf.png
Anya - Data is based from the real Anya, so it had a hard time balancing with the huge tits
https://files.catbox.moe/m4x9fp.png
https://files.catbox.moe/zm3bmr.png
https://files.catbox.moe/7w8064.png
https://files.catbox.moe/kyu8e5.png
https://files.catbox.moe/xgyt33.png
Ao - Now more girly and less handsome
https://files.catbox.moe/wqc3d1.png
https://files.catbox.moe/1ubbhh.png
https://files.catbox.moe/l4boe3.png
https://files.catbox.moe/mlp04l.png
https://files.catbox.moe/vzfiyc.png
Aqua - Beloved onion can finally have her well-deserved rest now
https://files.catbox.moe/v2ve6b.png
https://files.catbox.moe/b5z8f6.png
https://files.catbox.moe/n07bxs.png
https://files.catbox.moe/ie7130.png
https://files.catbox.moe/fqejux.png
AZKi - Programmed like a true diva, she only knows to keep singing
https://files.catbox.moe/kox2gp.png
https://files.catbox.moe/0xvddi.png
https://files.catbox.moe/bwet6y.png
https://files.catbox.moe/a1o842.png
https://files.catbox.moe/n14kuo.png
Bae - Mating presses may cause short circuiting but it is encouraged
https://files.catbox.moe/iqp4sh.png
https://files.catbox.moe/4iu5yw.png
https://files.catbox.moe/mqwmjy.png
https://files.catbox.moe/9mrcf6.png
https://files.catbox.moe/wc9mn4.png
Elizabeth - A queen must know how to mindlessly serve her people
https://files.catbox.moe/6bnfb5.png
https://files.catbox.moe/r89vo1.png
https://files.catbox.moe/pa1nej.png
https://files.catbox.moe/8a63fb.png
https://files.catbox.moe/kyqvmt.png
Fauna - Nature perfectly imitated and, to some, even surpassed by technology
https://files.catbox.moe/s7c86z.png
https://files.catbox.moe/rhawzk.png
https://files.catbox.moe/6c8fx0.png
https://files.catbox.moe/u8ddv6.png
https://files.catbox.moe/m7q11u.png
Flare - Being a robot and all, it's now more likely that she actually serves chocolate milk from you-know-what
https://files.catbox.moe/jletwy.png
https://files.catbox.moe/r825j8.png
https://files.catbox.moe/qdt18c.png
https://files.catbox.moe/1a4jqj.png
https://files.catbox.moe/jdp4ro.png
Fubuki - Not a robot friend, but a robot wife
https://files.catbox.moe/7yz77g.png
https://files.catbox.moe/so0xwl.png
https://files.catbox.moe/kaetk8.png
https://files.catbox.moe/2quedw.png
https://files.catbox.moe/fogdap.png
Ina - All that hardware is not just for looks as it is also dedicated to efficiently generating puns at any given time
https://files.catbox.moe/limf5e.png
https://files.catbox.moe/xou9c9.png
https://files.catbox.moe/fxpmuh.png
https://files.catbox.moe/qzmp62.png
https://files.catbox.moe/ggwk4m.png
Iofi - It is rumored that allowing paint to drip inside may cause her to switch to abstract art
https://files.catbox.moe/fvxaub.png
https://files.catbox.moe/trehsu.png
https://files.catbox.moe/23oeu4.png
https://files.catbox.moe/655leu.png
https://files.catbox.moe/dv36y6.png
Kaela - She is now the "help" that you get
https://files.catbox.moe/w5rwg1.png
https://files.catbox.moe/12pakq.png
https://files.catbox.moe/9dh9aq.png
https://files.catbox.moe/spsso0.png
https://files.catbox.moe/8l17pr.png
Kanade - Software is dedicated into being in different levels of smugness
https://files.catbox.moe/2h20kn.png
https://files.catbox.moe/k8csgf.png
https://files.catbox.moe/ag7nou.png
https://files.catbox.moe/gk01ji.png
https://files.catbox.moe/xakew9.png
Kanata - Same with Anya, but she can change her grip level
https://files.catbox.moe/jetnv0.png
https://files.catbox.moe/w7o63l.png
https://files.catbox.moe/y4qwqs.png
https://files.catbox.moe/uqwcvu.png
https://files.catbox.moe/oscdd1.png
Kiara - Artificial chicken never tasted better
https://files.catbox.moe/um3ewh.png
https://files.catbox.moe/gjvoxb.png
https://files.catbox.moe/wtrb2w.png
https://files.catbox.moe/jswg0b.png
https://files.catbox.moe/cv2zjb.png

>>83825615
2nd batch, with the next one being the last
Matsuri - Just add water for fireworks fitting for a festival
https://files.catbox.moe/oygn55.png
https://files.catbox.moe/356zmx.png
https://files.catbox.moe/2e33ta.png
https://files.catbox.moe/5ty35y.png
https://files.catbox.moe/d8efg0.png
Mio - Motherly mechanical Mio's massive majestic machine milkers
https://files.catbox.moe/oem6nj.png
https://files.catbox.moe/vkn6ck.png
https://files.catbox.moe/tte3rp.png
https://files.catbox.moe/jrc7ug.png
https://files.catbox.moe/kkaf29.png
Moona - You can see four moons on this mechanical moon
https://files.catbox.moe/ypm5zn.png
https://files.catbox.moe/yyqgrr.png
https://files.catbox.moe/pzlpsh.png
https://files.catbox.moe/wqfg78.png
https://files.catbox.moe/418xjn.png
Mumei - The Moom machine is the result of several centuries of advancement of civilization
https://files.catbox.moe/91866l.png
https://files.catbox.moe/zo3t4m.png
https://files.catbox.moe/kg66qh.png
https://files.catbox.moe/90q6n0.png
https://files.catbox.moe/cyczm9.png
Noel - A machine that shotas should avoid
https://files.catbox.moe/ej70oh.png
https://files.catbox.moe/nmf3g6.png
https://files.catbox.moe/x0xxce.png
https://files.catbox.moe/ynyqgr.png
https://files.catbox.moe/ldfxbr.png
Ollie - This robot may be based on a zombie, but this fembot doesn't have detachable body parts
https://files.catbox.moe/l1v991.png
https://files.catbox.moe/sz3v60.png
https://files.catbox.moe/hsv8ii.png
https://files.catbox.moe/fy92hj.png
https://files.catbox.moe/35mojn.png
Polka - Fembot fennec fox's furiously fat feminine floaters
https://files.catbox.moe/4xalis.png
https://files.catbox.moe/fgzs96.png
https://files.catbox.moe/6q991c.png
https://files.catbox.moe/tofzz5.png
https://files.catbox.moe/p8cng2.png
Raden - Rumored to be quite the "smoker" when wet
https://files.catbox.moe/5m1fbs.png
https://files.catbox.moe/8o6poh.png
https://files.catbox.moe/9l45ms.png
https://files.catbox.moe/554sgq.png
https://files.catbox.moe/6y9wn3.png
Raora - A robot this big and beautiful is certainly an art piece on its own
https://files.catbox.moe/4xnr4t.png
https://files.catbox.moe/ezrach.png
https://files.catbox.moe/bqpp1h.png
https://files.catbox.moe/cmyyg8.png
https://files.catbox.moe/wf9572.png
Reine - The turkey-tron 3000's tantalizing treasured tits
https://files.catbox.moe/y6dy6c.png
https://files.catbox.moe/mz1kxg.png
https://files.catbox.moe/u58q7e.png
https://files.catbox.moe/g0k294.png
https://files.catbox.moe/8duhr2.png

>>83825784
Last batch. Sorry for the spam.
Risu - Unlike the real Risu, it's not just nuts that she's after, but also bolts
https://files.catbox.moe/bm44bk.png
https://files.catbox.moe/gsjtkt.png
https://files.catbox.moe/1xebpp.png
https://files.catbox.moe/bjt53z.png
https://files.catbox.moe/k1u774.png
Sora - I think we already know if we force this android idol to drink a soda
https://files.catbox.moe/7mt27o.png
https://files.catbox.moe/vihcs1.png
https://files.catbox.moe/uk8167.png
https://files.catbox.moe/h78j8v.png
https://files.catbox.moe/90fozw.png
Zeta - The secret of this Zeta being robot is not even a secret at this point
https://files.catbox.moe/3pxfgc.png
https://files.catbox.moe/ru7e0m.png
https://files.catbox.moe/1hwwdu.png
https://files.catbox.moe/e2pt6c.png
https://files.catbox.moe/ubqdd9.png

>>83819294
Request Hikanari Hina walking in a fancy home setting while brgin either with a swimsuit, nude or both.

accidentally used lr 2e-3 instead of 2e-4 and lora still works fine. wtf?

>>83819294
Futa Amelia erection close-up, small dick and average dick variants

>>83821446
Seconding this

How do I get started doing this? I installed a few models just to see if it works. Generic prompts generate just fine but with the chuuba Loras I found, they were either not compatible or were not accurate at all.

Just Bote.
>>83826365
There are two "families" of models.
General XL models (usually based on Animagine for anime) and Pony.
LoRas made for one family will not work on the other (they can, and with some experience you can try them, but for a beginner stick to the correct ones).
Maybe you were using incorrect LoRas? Post a catbox of one of your gens so we can look.

>>83790227
This one actually become bigger that I planned for, because I miss read the request but Hey More for your request.
https://files.catbox.moe/gggk49.png
https://files.catbox.moe/pb0j8b.png
https://files.catbox.moe/1wku3q.png
https://files.catbox.moe/uvpg6d.png
https://files.catbox.moe/hnalp8.png
https://files.catbox.moe/tq2n7u.png
https://files.catbox.moe/j7mgxr.png
https://files.catbox.moe/2e6qpp.png
https://files.catbox.moe/jo2ydr.png
https://files.catbox.moe/ukq3tb.png
https://files.catbox.moe/gzvp1o.png
https://files.catbox.moe/5iqfq3.png
https://files.catbox.moe/171w18.png
Bonus Stage
https://files.catbox.moe/z9pmif.png
https://files.catbox.moe/u67seh.png

The extra is a Modern Goth Miosha
https://catbox.moe/ c/ s5ky00

>>83825615
>>83825784
>>83825821
Damn man, you didn't bake, you become the whole bakery

>>83806416
I brick my A1111, so I toke the chance to try comfy, is fun to use once you get use to it, but still has a little stiff learning curve.

Have a good night /vtai/

>>83826365
if you're using civit then you could see the base model for the Lora/checkpoint. Some Loras also have activation tags.
Best way for us to see what you're doing wrong is to post a catbox

>>83826480
Oh, I literally just downloaded it so I don't have much but that explanation makes a lot of sense. I'll keep that in mind and try to have better results in the morning.

>>83826567
Gothic Miosha is making me feel things.

>>83821446
https://files.catbox.moe/z79g0j.png

i compiled a folder a little more than double like 176 pics for from last time with more full bodies and it would make for a better model but i dont know how to make a good LaRA
i have the new . rar with all the files for those that would help

>>83759052
GEEGA coomer back again asking for help
i compiled a folder a little more than double like 176 pics for from last time with more full bodies and it would make for a better LaRA but i dont know how to make it myself
its too many for a cat box galley but i have a compressed rar with all the images i feel someone here would know better what to do with it to make a good LaRA
>it was too big for just catbox sorry
https://litter.catbox.moe/vl74yg.rar

>"U-Ueh? No way! You painted my nails while I was asleep? Thank you Friend!"

>>83823159
Thank you anon! NAI is great at doing feet but (at least for me) it struggles with detailed backgrounds a lot. It makes sense though, the further things are from the center of the screen, the more the AI will struggle generating fine details, since it will prioritize the objects in the center first. The struggle made me appreciate genned backgrounds even more, and yours is pretty cute!
>>83822201
It's pretty crazy how useful knowing how to get around photoshop can be for your gens. I started using editing tools too to fix some tiny imperfections or to add new stuff to the gens and it helped a lot. Great job with this one! cute moom
>>83825615
>>83825784
>>83825821
Damn, anon... And here I am, struggling to put out 1 or 2 gens at a time, lol. Maybe you should do a .zip of all of these gens? Just in case someone wants to download them. I've... been on the requester end of downloading 20 individual pics one by one, it's painful...
>>83826648
Keep it up anon, good luck getting everything ready. I still remember some people posting their first local gens who later became regulars. I wonder what kind of gens will you post when you get your setup up and running?

>>83819294
Matsukai Mao swimsuit getting revealing her inverted nipples

>>83819294
Anya buns and thighs

Good morning /vtai/. Haven't been posting for the past few threads so sorry if I missed anything. Been busy with playing games. I see some requests I can fulfill so that's nice, I'll try to get them done.

>>83819663
Hello Pekora-anon. If you're available/don't have anything to prompt then, care to fulfill a request of mine?

>>83825403
Rest well, TWLAnon.

>>83826480
I love how big and fluffy her ears turned out to be in this style, could you please post a catbox?

What the fuck is this "Early Access" dogshit on Civitai? What a crock of shit.
>>83829301
https://litter.catbox.moe/cfapf0.jpg

>>83828409
The rar is already gone. You can use gofile. What was wrong with the first one other than half the images not loading?

>>83829739
its mostly that its very stiff with expressions i used to many screencaps of Geega's face and also not enough full body references that are there in this
the biggest problem is with it forcing the model's art style and expression, the eyes being aimless in nearly all gens, and it leading to artifacts if you try to use other loras with it.

>>83828867
My mix kept trying to white-wash Mao... In any case, Mao nipples
>https://postimg.cc/gallery/mZVLBJD
>>83829301
Sure, even if I was busy prompting something else I'd still save your request for later, don't worry. Hope you had fun these past days away from /vtai/

>>83829889
>>83829739
i went to Gofile to upload the rar here you go
https://gofile.io/d/pew9NX

>>83826567
>Modern Goth Miosha
holy sexo

>>83829889
If I train it on civitai again, it will have the same issues. have you tried the epoch 7 one? There were two files in my original post.

>>83830614
i can do that and see if there is more of a difference

>>83829364
Thank you very much. What's that model? I can't find it anywhere, it's a personal mix of yours?

>>83830025
>My mix kept trying to white-wash Mao
NTA, I don't know which tags you used but try "tanned/dark-skinned female". But I assume you used these, in which case could be a model issue.

>Sure, even if I was busy prompting something else
Oh, sorry. Take your time with my request then, I don't want to rush ahead of your already planned schedule.

>I'd still save your request for later, don't worry.
Thank you. In any case my request is pretty simple - pov Botan blowjob in her first outfit. Background being simple "indoors, living room" type of deal. That being said, please make the guy light skinned, just like you did in Tsuna set. I assume you were going to prompt that regardless, but I wanted to clarify that just in case. Thank you very much in advance.

Forgot to ask you, have you settled on your new style yet?

>>83830682
There should be. I shared it exactly because after epoch 7 I have seen a big shift in the samples it was throwing out.

>>83830696
It's a mix based on Autism Confetti, a bunch of models added on, restored the original Text Encoder and baked in HLL, Fumio and NatThe Lich LoRas.

>>83828637
It's my first time zipping it like this. I'm pretty sure you just rename each file under one name then open up the zip file then extract everything.
https://files.catbox.moe/q23oct.zip
https://files.catbox.moe/doukrb.z01
https://files.catbox.moe/1cplp4.z02
https://files.catbox.moe/ei2zyu.z03
>And here I am, struggling to put out 1 or 2 gens at a time, lol
Well, there were days of just waiting and waiting

>>83830696
Sure! Give me a moment. Just a heads up, my new mix is kinda cute...
>have you settled on your new style yet?
Kinda, I'm still testing a lot of different artists. I have a danbooru tag compilation link and every day I set aside some of my free time to check through it to find new styles. Sometimes I find unexpected styles like picrel. I have an 8+ artist mix for feet that it's starting to grow on me, so maybe I'll settle on it, though I doubt it.
>>83831037
Honestly, I just make 1 big .zip and post the whole thing. I separate the gens in folders inside that .zip, but catbox does the trick too. If you see me using gofile+postimg is because my ISP has catbox banned and making catbox links/albums is a pain (though doable). Fun tip, if catbox+litter are down (literally an everyday occurence) postimg is an alternative to them. I learned it from one of the thread's OP links.

Got this lora a while back after seeing it in the thread but never got around to trying it
https://files.catbox.moe/qlzqwo.png
https://files.catbox.moe/fbznuy.png
https://files.catbox.moe/i5i83n.png

>>83830696
Here you go! I tried to make you a sequence, I hope you like it.
>https://postimg.cc/gallery/JJ8JCz8
Ignore the minor inconsistencies such as the guy's clothing, NAI sucks at doing sequences, (at least for me)... You were really kind when I first started posting here and you're still very polite to this day, so I went through the trouble of enhancing every single pic of the set. The 4th one I couldn't enhance do to issues with colorize, so if you notice a drop in quality, there's that. I'm going to take a break, but I'll be around here later in the day. See ya, /vtai/

>>83819294
Re-requesting >>83791991
>Requesting doggystyle analRyS, with her ass cheeks spanked, hair being pulled and doing a tongue out ahegao.
Include an aftersex shot of her laying down the floor/bed and cum flowing out of her asshole if possible.

>>83819294
Throwback Tuesday, requesting some old school chinese prompt alchemy (https://docs.qq.com/doc/DWGh4QnZBVlJYRkly)), used on newer models. Make it a Reine or something

>>83817764
Thank you for the goods

>>83832468
Unprompted prequel:
https://files.catbox.moe/qeqttg.png
A few attempts:
https://files.catbox.moe/1wq1du.png
https://files.catbox.moe/xivn69.png
https://files.catbox.moe/1810yh.png
https://files.catbox.moe/566fdd.png
https://files.catbox.moe/iveo2q.png
Aftermath:
https://files.catbox.moe/34smfk.png
https://files.catbox.moe/bu3v9p.png

>>83822060
This is an old set but I hope its to your satisfactory. I added the bj as a bonus.
https://files.catbox.moe/m2ozl9.PNG
https://files.catbox.moe/6x5okc.PNG
https://files.catbox.moe/p39wsw.PNG
https://files.catbox.moe/dzx8b2.PNG
https://files.catbox.moe/glcgcc.PNG
https://files.catbox.moe/wj17en.PNG
https://files.catbox.moe/ydz7nc.PNG
https://files.catbox.moe/vcvhoc.PNG
https://files.catbox.moe/to71qf.PNG
https://files.catbox.moe/m7wtbv.PNG

Rest of the set.
https://files.catbox.moe/095tl7.zip

>>83830782
I see. Well, looks like I am out of luck trying this model for myself then. Unless you're willing to share that model, but I won't push you if you don't want to.

>>83830025
Whoops forgot to reply to that last line.

>Hope you had fun these past days away from /vtai/
I did have fun but I also don't like missing out on stuff, that also includes me being here and seeing what other anons prompt/sharing my own prompts/doing requests. But there's a positive that I am hit with a lot of inspiration for my newfound ideas, I basically have a full checklist of what I want to prompt in the future lol.

>>83831482
>Sure! Give me a moment.
I'm not going anywhere, don't worry.

>Just a heads up, my new mix is kinda cute...
I assume that's the mix you currently use/used for Tsuna & Mao sets? If so, I want Botan in those too. It's a very pretty style.

>Kinda, I'm still testing a lot of different artists.
I see. No rush. I'm sure you'll find something you can set on. I am also experimenting here and then with artists I already love, but haven't really tried myself to mix them.

>I have a danbooru tag compilation link and every day I set aside some of my free time to check through it to find new styles.
That sounds really troublesome. I guess you don't check individually how each artist looks like and just put them in a blender to prompt and see how it looks?

>Sometimes I find unexpected styles like picrel.
It's uhh... Uncanny, if I may be so honest. I've never liked the semi-realistic or realistic styles.

>I have an 8+ artist mix for feet that it's starting to grow on me, so maybe I'll settle on it, though I doubt it.
That's good to hear! I'm glad you found something specific for feet. If you need recommendations, try casino's style. He's very oriented on feet, maybe it could be for you. Well even going outside of our AI realm, his art is beautiful even to me, someone who isn't into feet at all, I appreciate his work.
That being said, I wish I could find a mix of artists that could replicate the perfect midriffs of SD1.5/heavenorangemix/creamsicle days too.

>>83832233
>Here you go!
Oh, that was so fast, thank you very much! I was still typing my reply to your previous post and you already went and completed my request. Very cute picrel Botan too! Her smile is the brightest and heals my heartbroken soul, much appreciated.

>spoiler
That's fine. You could experiment with "clothed_female_nude_male" tag if you so desire. I'm confident your current mix should be able to prompt that.

>You were really kind when I first started posting here and you're still very polite to this day
I try my best to be kind to every person who is also equal in kindness. "Treat others how you want to be treated" as they say. Even if we're strangers who have met on taiwanese basketweaving forum dedicated to prompting AI artworks of virtual youtubers, it doesn't mean we shouldn't be polite to each other.
Naturally, I cannot be kind to everyone and there are certain people whose actions who ignite the most raging inferno of fury within me, and I can't help but lash out from their misdoings. Sorry, I didn't mean to go on a tangent about myself that nobody should really care about.

>so I went through the trouble of enhancing every single pic of the set
I don't know what exactly what that process was or which parts exactly you enhanced... I assume inpainting some parts? Either way, I love how cute she looks in the first pic of the set, the catface with the tongue with the fang is adorable.

>The 4th one I couldn't enhance, so if you notice a drop in quality, there's that.
Honestly, it's very hard to spot the difference and if you didn't point out that it's the 4th, I wouldn't have noticed. I can see the that detail on the hair a bit different and perhaps she has less details on the eyes? Can't notice anything else.

I hope you don't mind this criticism, but I'll point out a different problem in that set - is that the 2nd gen is very loli-ish. I assume that's the issue of not having any "breast tags" so it would default to a certain bodytype your style is biased to. The 4th gen also shares a bit of this problem with her shoulders to head ratio being very small, but not as offensive as the 2nd.

Other than that problem, I love this set! It's a very nice sequence and I'd definitely ask for more when you have the time to prompt.

>I'm going to take a break
Take care, Pekora-anon. Enjoy your rest and have a good day.

>>83827419
NTA but this is super cute

Testing.

>>83822201
Nice, I thought getting more than two people was hard

>>83819294
Reinstating from a two threads back:

Requesting a bedridden Marine feeling pleasure after getting a catheter, making the nurse rather uncomfortable.

>>83834895

>>83834895
>>83835911
I think your test was successful

>>83819294
seductive gothic vampire kronii

>>83835911
>>83836074
I'm trying out a recent model.
It seems decent enough at hands. Prompt adherence and backgrounds aren't bad either for Pony.
It absolutely does not give a singular, lonely flying fuck what style LoRas I tell it to use though. Character ones work fine.

>>83825615
>https://files.catbox.moe/9mrcf6.png
>bae bot rat bot sex got
spoken like a true brrat

>>83836899
peropero

>>83836899
TWL!? I-Is that y-you?

>>83819294
Shiori choking Missionary sex with her passing out right after climax.

>>83836997
Yeah.
>>83837006
No it's me, Anon.

>>83821394
This took some effort, it's not perfect but I hope you'll like it. I can't pick which ones are the best so you can have the alternatives as well.
https://files.catbox.moe/lb2851.png
https://files.catbox.moe/ynernk.png
https://files.catbox.moe/4jce06.png
https://files.catbox.moe/621rb7.png
https://files.catbox.moe/l6rns6.png
https://files.catbox.moe/ukb16j.png

>>83835911
Very erotic cat.

PonyColors is out. dampened a bit the level of colors I had in "AutisticBackgrounds" to prevent frying.
https://civitai.com/models/692446

Picrel for lack of actual vtuber images on it without loras because pony is shit and don't have time to try tags. Gotta run. Ta-ta

>>83824757
It does seem to help in Flux to help aim for a style since it doesn't really use artist tags and stuff for style control. Though I haven't done extensive testing on it.

>>83826060
Kek, still would.
>>83832133
SEXO!
>>83828637
Cute Moom feet, looks soft.
>>83833578
Absolutely incredible Pekomums
>>83836483
The colors really vibrant so that's a plus. Especially like the Marine one.
>>83837288
I can't believe it's the anon!! Joking aside very cute miko (and sideboob)
>>83837507
Thanks will take it for a run. Cya later

>>83838172
why is she so fidgety

>>83838541
Keeping that weight balanced requires constant adjustment

>>83836483
Some styles work. Somewhat.
>>83837988
Marine was with the Asanagi LoRa.
It definitely did something. The tits, mostly.

>>83824714

>>83837386
Whoa damn nice. She is taking it like a champ

>>83773035

>>83667773
Slight necro. I have discovered that I have absolutely no self-control, and this sequence reflects my downward spiral. Adetailer fucked up some gens - pretend they are lipstick smudges or something.
https://catbox.moe/c/nypfdi
https://files.catbox.moe/x0oosu.zip

>>83777774
Thanks Anon for the kind offer. Let me scour twitter and booru one more time and take some of the outfit reveal screenshots, and get back to you within the next day.
>>83822201
Amazing work TWL, and on >>83808169 too. I should start practicing regional prompts. Thanks for testing the guided penetration LoRAs too, picked up Hoseki.
>>83825615
>>83825784
>>83825821
Holy shit Anon you went on a bender. I hope you've been hydrating, what a fever.
>>83826567
Damn, that is sexy, modern goth Miosha activating my neurons, giving me inspiration to try this out on other chuubas.
>>83832133
Great minds think alike, but which LoRA did you use? I had a hell of a time trying to wrangle this view between legs LoRA.
>>83836899
>>83837288
Nice eyes and sideboob.
>>83837507
Thanks for the model, will take it for a spin too.

>>83839338
>which LoRA did you use
This one https://civitai.com/models/325951/sword-swallowing-position-throat-fuck-or-pony-xl-concept

>>83825403
Requester here, thank you. Now Aqua is ready to go

>>83839124

>>83832507

>>83833578
Never saw AI do an unpeeled hard penis, looks quite good with Pekomama sucking only the tip

>>83838541
she's ready for this 526469

>>83839338
SEXO

Good morning /vtai/ :)

>>83819294
Biboo Drinking Strawberry Milk, in a BDSM situation

>>83840034

>>83841126

>>83777774

Oh that actually went faster than expected; there's really not that much art out there. I think some screenshots are a little blurry, sorry! Let me know if there's anything else I can do on the dataset.

https://gofile.io/d/Lgmq6C

>>83841165
Damn these chinese alchemy stuff still work pretty great. These are beautiful

>>83840979
rare spice gen

>>83839912
WARNING
PENIS BEYOND THIS POINT
https://litter.catbox.moe/r48uym.jpg

>>83841766
Me.

>>83841018
>>83841152
lovely pits

>>83819294
Futa Advent and Justice comparing dicks(very large)

>>83841250
Thank you, that should be good to get me started. I'll try to have it tomorrow or next day

>>83841447
>>83841512
I kneel to these tits!
>>83841766
Holy sexo!

>there's no lora of watame's burlesque-like outfit from the swing MV, only the red/black dress she wore during live concert
LIFE IS PAIN
FUCK

>>83841766
She will die...

>>83842504
These are actually the golden proportions

>>83842632
Pretty sure someone here made one before

>>83843603
Please god tell me that you still have a link or that it's easily findable in the archives, I crave sheep sex

>>83819294
Hags deflowering with a rather embarrassed but happy expression.
Any hag chuuba will do.

>>83836493
That last one was actually supposed to "Bot" too but somehow the upscaler changed it into a "G" and I didn't realize it
>>83839338
Can't hydrate when I'm busy pouring the water onto these bots

>>83840154
Foreskins are underappreciated, aren't they?
https://files.catbox.moe/vmwgqy.png
https://files.catbox.moe/ydwrsi.png
https://files.catbox.moe/ag49wy.png
https://files.catbox.moe/mzlpo0.png
https://files.catbox.moe/tqfy6d.png
https://files.catbox.moe/mqnjuf.png
https://files.catbox.moe/7c6u0o.png
https://files.catbox.moe/mz5r75.png
https://files.catbox.moe/yk5lfp.png
https://files.catbox.moe/vgrpau.png

>>83843765

>>69204680

>>83841766
No penis here. Just nude rabbit.
https://litter.catbox.moe/hpo0p5.jpg

>>83844855
very nice style

>>83844218
>https://files.catbox.moe/mz5r75.png
This one is my fav

>>83819294
Koyori drinks/feeds one of her genmates a potent fat nipple potion

>>83819294
Preggo Mio

>>83844251
>read through the thread
>everybody uses comfyui
I'm too stupid for this, it's over.

>>83844855
And PekoMother.
https://litter.catbox.moe/lcdohj.jpg

>>83847165
Wouldn't it be funny if PekoMama had a big ol dong? I think that'd be a good joke

>>83819294
Requesting intense shower sex with Scarle

>>83846623
https://files.catbox.moe/0ribyu.png
https://files.catbox.moe/eyovpd.png
https://files.catbox.moe/wrzx4x.png
https://files.catbox.moe/j9th0r.png

>>83844218
Good lord, this is a fantastic set

>>83839338

Ass to ass
https://files.catbox.moe/yijz4j.png

>>83839563
Thanks, maybe this one will play better.
>>83840034
Lovely painting style.
>>83844218
Foreskin insertion is a great tag. Are you also using the LoRAs? Haven't tried it with base tags.
>>83844855
>>83847165
Interesting model and nice style roulette. Incase is always nice to see. Where did you get the Nagu LoRA?

OL anya in different styles
https://files.catbox.moe/tzl5w6.png
https://files.catbox.moe/60zhdc.png
https://files.catbox.moe/zz2ro0.png
https://files.catbox.moe/oxbpr7.png
https://files.catbox.moe/ag8dz4.png
https://files.catbox.moe/io4w03.png
https://files.catbox.moe/ayj0q3.png

>>83848528
was expecting office lady but this works too, why does she have a furry nose

>>83847996
>Where did you get the Nagu LoRA?
I think it was from a Mega by someone here.
Here, I uploaded it. I tried catbox but it kept failing.
https://gofile.io/d/LgiOoU

>>83819253
who is this chubba name? I saw someone posted it here but i forgot her name.

>>83819294
>chuubas whose LoRA you downloaded solely based on their design without knowing anything else about them
Go.

>>83849144
I believe that is Whisky project

Has anyone gen'd Gura and Panko seggs yet? I want to see them kissing eachothers mankos.

>>83849298
I used to watch her years ago before she ceased activities.

>>83849327
This but Gura stuffing Panko's manko with her double shark dicks

>>83849352
She literally stream a few days ago

>>83849621
Yeah, on Twitch. With no VODs. Might as well be dead.

>>83848636
dunno

>>83837988
>Absolutely incredible Pekomums
Glad you liked them, thanks.

>>83839262
Neat, thank you. Is this flux?

>>83840154
That's because nobody prompts it. One day, I wanted to experiment if it's something you could prompt and it definitely is.
https://files.catbox.moe/ijsgqu.PNG
https://files.catbox.moe/xf6wdj.PNG

Case in point, what the other anon who replied to you and made the set with it being the focus.

>>83844218
Very good job, anon. I love this Coco set, especially that last gen. I tried to prompt "tongue in foreskin" but never got any good results.

>>83839334
That's Enna for you.

>>83846720
Don't lose heart anon. I heard that there are forge users which is an equivalent of A1111. Then again I'm a poorfag who uses ai gen sites to make my prompts, so you can join me in my poorfag path, but if you have good pc then obviously learning how to set up local is in your best interest.

>>83846720
Only ever used Comfy for merging. Fuck using it to gen anything.
Get ReForged (it's Forge but forked so avoid the insanity going on there, and it receives updates)

>>83819294
Futa Pekora sex with Pekomama

Afternoon.

>>83850638
>Is this flux?
No, just SDXL (ponymagine+albedo mix)

>>83850638
Forgot how much I love Coco's fat dragon tits.

>>83819294

Requesting Park Dona sitting in your lap, and you are fondling her tits

>>83847249
Haha, what a goofy idea!
https://files.catbox.moe/6c64gz.png
https://files.catbox.moe/e091zf.png
https://files.catbox.moe/cgxpgi.png
https://files.catbox.moe/imdpkz.png
https://files.catbox.moe/wut45q.png
https://files.catbox.moe/21kw82.png
https://files.catbox.moe/hwbwhv.png
https://files.catbox.moe/sqn4ep.png
https://files.catbox.moe/idydby.png
https://files.catbox.moe/2gysg7.png
https://files.catbox.moe/hh0dek.png

>>83819294
Gura bukkake, gallons of cum.

>>83852919
Oh my how funny, what a splendid joke!

>>83852548
Nahone.

>>83852665
You and me both, anon. Her majestic tits are always a treat for sore eyes. The dragon trio is blessed in general. Coco, Elira, Selen. All of them share big personalities.

>>83852548
This is with flux base gen

>>83847996
Whew~ I am more into this than I thought I would be
>>83848528
>https://files.catbox.moe/ag8dz4.png
I like this one the most
>>83851909
Good afternoon! Thicc idol love
>>83854960
The dragon is tempting me
>The dragon trio is blessed in general
Kek that's true, Elira and Selen were hiding their power levels till the 2nd outfit
>>83855117
That's really good and she is cute there,
Could I get a catbox of the base gen if you still have it? Still trying to figure out a good prompt structure for flux.

>>83855790
Nice sidesack

>>83856728
Kek totally missed that. I'll pretend it is a minotaur speedo

Hangers.

>>83759484
I also did a couple of the other portraits
https://files.catbox.moe/1skpv3.png
https://files.catbox.moe/6dop03.png

>>83858493
Oh man these are great, thanks!

>>83849181
I made the LoRA cause someone here asked me to, had no idea a aerospace vtuber existed kek. I am now a huge fan of her,

Btw who is the chuba in your gen? She is really cute, honestly doesn't seem like the usual vtuber-esqe design, I really like it.

>>83858493
>Circle game
Damn I looked
>https://files.catbox.moe/6dop03.png
This is really good

>>83847996

>>83847996
>Foreskin insertion is a great tag. Are you also using the LoRAs?
Yeah I just figured I'd check for foreskin-related loras and found this one https://civitai.com/models/48999/foreskin-insertion
>>83850638
> I tried to prompt "tongue in foreskin" but never got any good results.
See above. I did have to do a bit of rerolling and editing but not a lot.

>>83847742
Glad you liked it! Have some more.
https://files.catbox.moe/d3ksy3.png
https://files.catbox.moe/8ci2as.png
https://files.catbox.moe/fqa2s3.png
https://files.catbox.moe/lp92qh.png
https://files.catbox.moe/npygte.png
https://files.catbox.moe/coxir7.png
https://files.catbox.moe/1gwq7d.png
https://files.catbox.moe/4q4q6y.png
https://files.catbox.moe/kg7o2t.png
https://files.catbox.moe/jucncc.png
https://files.catbox.moe/v13yd3.png
https://files.catbox.moe/ftzsjb.png
https://files.catbox.moe/yk6uhv.png
https://files.catbox.moe/bdrb7q.png
https://files.catbox.moe/h7p9yt.png
https://files.catbox.moe/yrph9t.png
https://files.catbox.moe/44dqnq.png

>>83859534
Her name is Sawa Grape Sama, apparently. Who's your aerospace chuuba? Sounds cool.

managed to make a (hopefully) better LoRA of GEEGA
to try to get more people generating more porn of her
>make more porn of her please
https://civitai.com/models/689651

>>83860661
Thank you cute name too.
>Aerospace Chuba
Usui Clear, Japanese Indie

>>83861461
Thanks for the LoRA. Those sample pics are looking good.

Still can't believe she's graduating today...

>>83825403
Evening folks

>>83713370
https://files.catbox.moe/xr0pls.png
https://files.catbox.moe/5z7ol6.png
https://files.catbox.moe/3j6vkr.png
https://files.catbox.moe/u3ucqo.png

>>83825615
Holy hell I was wondering where you gone but you were building a factory!
You pretty much did everyone, had to check out the Fauna and Raora ones, they look great
>>83826567
Nice set, Mio looking very pretty with the lace and frills
>>83828637
>It's pretty crazy how useful knowing how to get around photoshop can be for your gens
I don't really do anything crazy, I just gen large batches and Frankenstein something together.
Going to https://www.remove.bg/ really helps to just cut a character out from one gen and paste them to another
Using photoshop's content aware fill helps alot to just fill in empty space so when you img2img it has something to work with
Here is the Photoshop version before running through img2img inpaint
https://files.catbox.moe/34r8jo.png
Started with an IRyS boob suck then a generated a bunch of Mumei boob suck and picked one that I think would fit and transplanted her there.
It does take a lot of time though often an hour or two from start to finish
>>83832133
Oh the sword swallow LoRA, that was like the first LoRA I used when I made the switch to pony, can be a bit hit or miss given the head is tiled all the way back.
>>83833224
That's a nice gooy nephilim hole
>>83835281
Still is, this took a little under 2 hours to build and blend together. 2 girls isn't bad but 3 girls will need some major work either in photoshop and in img2img inpainting. Lots of iterations to make everything look seamless.
>>83837006
kek no, looks like BibooDressUp anon?
>>83837507
I'll take a looks and see if I can merge it with something
>>83838172
Hello OL Anya, beeg as always
>>83839338
Wow you killing it with these sets, so much variety and expressions
>>83840979
Morning, Shiori not cold there, she is even sweating a bit
>>83844218
Leave it to the furries to have several foreskin tags
>>83844855
Oh that is a very pretty Peko body
>>83846720
There are still people here that don't use ComfyUI, I'm too lazy to change and/or spend a week trying to figure it all out.
>>83851909
Good afternoon
>>83852548
Pretty good skateboard gen, what's the ratio of that mix? Albedo is pretty knowledgeable

>>83862704
Man right in the feels...

Could I get some IRyS please?

>>83819294
requesting Shion drinking vodka with a straw

Don't die now

>>83863401
Noel being supportive, you love to see it

>>83819294
two HoloEN members of your choice, wearing bunny suits, getting bukkake'd

>>83852919
This is so damn hot. Would love to see more busty futa Holos from you!

>>83860641
You spoil me anon, thank you very much for these.

>>83819294
Requesting that YOU add "human centipede" to your gen.
https://litter.catbox.moe/y5edx1.png

>>83863401

>>83723045
https://files.catbox.moe/ar3tbo.png
https://files.catbox.moe/sce06l.png

>>83861461
have this in the meantime

>>83869421
Is that hades style? Looks cool

>>83861461
>to try to get more people generating more porn of her
isn't this opposite of what geega wanted? kek

>>83855790
>Kek that's true, Elira and Selen were hiding their power levels till the 2nd outfit
Selen? Yes, definitely. But Elira? The hints were there from the start.
But ultimately Elira won in that battle because she got not 1 but 2 pure sex outfits. (Not counting the default). Selen only got 1 great one and its the second, the rest were a downgrade.

>>83857967
That top is hanging onto it's life.

>>83860641
>See above. I did have to do a bit of rerolling and editing but not a lot.
Huh, I see. I didn't know there was a lora for it. I'll try it out sometime, thanks. The gens I posted were all done without any loras.
Also nice set continuation.

>>83863401
Good evening, TWLAnon.

>live noel reaction
Amazing.

>>83871284
Nobody can stop the process of horny people getting work done. Just look what Suisei's reverse psychology did to her R18 art numbers. She's one of the most lewded chuuba despite never being close to fans or actively trying to pursue the career of "please make more lewd stuff of me". If anything, it can backfire - Nene.

>>83869169

>>83694544
https://files.catbox.moe/8c1zgn.png
https://files.catbox.moe/pa4lvu.png
https://files.catbox.moe/pp2v7z.png
https://files.catbox.moe/7mf01x.png
https://files.catbox.moe/mlzhde.png
https://files.catbox.moe/o6q29l.png
https://files.catbox.moe/qbfr0a.png
https://files.catbox.moe/vqohv3.png
https://files.catbox.moe/2qyyz7.png
https://files.catbox.moe/wyuxpl.png

>>83872226
Hello,
>live noel reaction
Nothing a little bit of photoshop can't solve

>>83873545
I am really liking this style, vibrant and the eyes really pop

>>83853272
I made your usual Gura facial bukkake I but also took the opportunity to use your request as an excuse to whip out *the cumbath* prompt and make some separate gens just for fun (Imagine giving Gura a bukkake so big that you're able to fill an entire bathtub...) I couldn't delete Bloop from one of the bathtub pics so try to imagine he's a rubber duck or that he's very, VERY loyal to Gura...
>https://postimg.cc/gallery/qJ6Qtrx

>>83843778
Deflowering is a very niche kink which I coincidentially also have. I think I've never seen it around here, either requested or genned. I think that in the context you requested it can be quite cute/lovey-dovey. Since any hag chuuba was ok, I picked Botan
>https://i.postimg.cc/1m7WqPQh/Shishiro-Botan-deflowering.png

>>83833578
I don't know if this applies to Local, but for NAI, on their wiki pages, Danbooru(Danbooru tags were used to train the NAI model) has a tag_groups link which have different artworks with the same topic(tag). If you want your mix to be good at something specifically (for example, I wanted to focus on feet), you can search that tag_group and look up the artworks. Your artist suggestions aren't bad by any means, in fact they are quite popular, but I don't personally like to use them unless I want the gens to imitate their style because even with less powerful brackets their style is quite overwhelming and easily recognizable. I had fun with the Asanagi style once, though it needed some tuning. Also yeah that "realistic" style can be quite uncanny sometimes, not planning to use it, I just found it amusing. Can make cool stuff though!
>Enhance
NAI has a feature that upscales your image and runs your prompts multiple times to "upgrade" the quality of the pic. It smoothes things out, which sometimes helps, but can erase some fine details like lineart in the process. Since it upscales the gen, it costs tokens to use. The higher the strength of the enhancing the more expensive it is, though I find such strength values unnecessary. I like using the minimum ones sometimes. Makes the gen very smooth-looking and fixes a lot of imperfections. Im aware that you can enhance without upscaling but I don't think it's worth it.
>the 2nd gen is very loli-ish
I do love to use the "cute" tag for everything so I can see that happening, lol. Either way, I'm very glad you enjoyed the set!

>>83871284
Very cute Selen!

>>83853272

https://files.catbox.moe/o79car.png

>>83867833
https://files.catbox.moe/cyo67b.png
https://files.catbox.moe/mq34ma.png

>>83873545

>>83725717
https://files.catbox.moe/7uhax5.png
https://files.catbox.moe/mes72o.png
https://files.catbox.moe/7xbloc.png
https://files.catbox.moe/nqoo68.png
https://files.catbox.moe/fbo0hv.png
https://files.catbox.moe/s1cwhd.png

>>83878678
Damn didn't think AI could get that much glue, I assume you needed to lean on specific artist or did you inpaint a ton of white stuff

>>83876021
>>83878678
Nice Gooras. kek that Bloop.

>>83819294
kanata breakdancing at the olympics

>>83819294
Re-requesting more options for korone with nothing on but her panties with a smug expression on her face.

>>83880240

>>83745749
Taking a page from Anya poster
https://files.catbox.moe/2fh3yq.png
https://files.catbox.moe/kpmd6f.png
https://files.catbox.moe/w30e09.png
https://files.catbox.moe/q0dv3p.png
https://files.catbox.moe/hqoqph.png
https://files.catbox.moe/4az6qj.png
https://files.catbox.moe/2zgwe8.png
https://files.catbox.moe/jca9by.png
https://files.catbox.moe/vutqj3.png
https://files.catbox.moe/b5vz18.png

>>83884649
>https://files.catbox.moe/b5vz18.png
WHOA

>>83826567
>I brick my A1111, so I toke the chance to try comfy
Man, I bricked my A1111 and I think my Forge is going the same way too. Maybe I'll force myself to do the switch too...
Good night!

>>83816916
Wow, thanks my guy! I figured this would get buried since it's somewhat obscure and potentially hard to gen but thanks for the crack at it!

>>83808803
Good evening and good night, /vtai/!
Weird stuff, pc-kun currently has a 50/50 chance of turning off whenever I run a batch larger than 6. It is relatively new, it shouldn't be like this. I'll just quietly bweh.
>>83813672
Cheers! GET MOTIVATED RAAAAAHHHH
>>83817063
Hello!
That other gen used a new(ish?) XL model, it's ponyai v6!

>"N-N-N-No touchy the crown! Don't!"

>>83867833
https://files.catbox.moe/aakxzr.png

>>83884649
Holy moly!
>>83885503
I haven't tried it yet but they have a new front end for comfy so I think you don't have to mess with the noodles much if you have that.
>>83885663
Extremely beautiful gen, gn!
>>83886335
But she seems so excited though!

>>83886746
Kek I just realized it said EN. my reading reps stink

>>83819294
HoloAdvent in a "The Usual Suspects"-style mugshot lineup

>>83886833
>>83886746
It's fine, the image was good anyways!

>>83878678
thank you!

>>83886746
>"GIVE IT BACK!"

>>83888663
Extreme gore warning

>>83823159
uuu trying out flux is tempting.
>>83822612
>>83825403
ah, an animagine lora. no wonder i thought they were sexo.
>>83857967
sexo
>>83862704
how was aqua's graduation stream?

>>83889466
>graduation stream
oh wait the live is later

Bae rejects

>>83863401
>I'll take a looks and see if I can merge it with something
If you use comfy for merging I'd recommend looking at the method used and just replacing pony with your base model instead. It will work better than merging this pre-merged model with another model and diluting the process.

>>83728392
https://files.catbox.moe/gd85xp.png
https://files.catbox.moe/9d0d2g.png
https://files.catbox.moe/8qdk9t.png
https://files.catbox.moe/t8iw7i.png
https://files.catbox.moe/ystbl0.png
https://files.catbox.moe/umcrwo.png
https://files.catbox.moe/k263lr.png

>>83889466
It's that plus some Wada Arco.

>>83884649

>>83756894
https://files.catbox.moe/tpsny5.png
https://files.catbox.moe/jnc0xc.png
https://files.catbox.moe/3d35vd.png
https://files.catbox.moe/s0cqab.png
https://files.catbox.moe/9ylalz.png
https://files.catbox.moe/4o7ozm.png
https://files.catbox.moe/tpu21c.png

>>83885663
Evening and good night BibooAnon, hope PC isn't dying.
>>83886335
oh no if take off the hat the Holos die!
>>83886746
Cute heart hands on that bukkake
>>83890605
>comfy for merging
I'm using SuperMerger maybe there is some block weight merge

>>83880240
Nice. Thankyou.

>>83891567
I merged the difference between HoloDayo and Animagine into the Out8 block. I'm not sure whether supermerger allows doing Model A - Model B merged into a specific block of Model C

>>83819294
Fuck it oppai loli-fy Scarle

>>83891402
that's actually what i had kind of assumed, something in that space. I was pretty sleepy when I saw them at first.

>>83891994
That's really cool! Specially that trail going off in the background and the sun being so bright yet only illuminating her backside instead of her front. As if saying that despite being in the foreground Usui is not necessarily the main subject.

>>83891989
Very cute image!
>>83891994
Wow, I really like this one

>>83687339
mao
https://files.catbox.moe/s9fqto.png
https://files.catbox.moe/tljnpc.png
https://files.catbox.moe/wx1sjf.png
https://files.catbox.moe/s7thjw.png
https://files.catbox.moe/fooz60.png
https://files.catbox.moe/42ie45.png
https://files.catbox.moe/75xjmm.png
https://files.catbox.moe/ct6oh8.png

guuba
https://files.catbox.moe/81x7g6.png
https://files.catbox.moe/8tfijs.png
https://files.catbox.moe/wgewu7.png
https://files.catbox.moe/hc55u1.png
https://files.catbox.moe/3yrby5.png
https://files.catbox.moe/0q5ckr.png
https://files.catbox.moe/40alrx.png
https://files.catbox.moe/y9ym9b.png

>>83891567
why is she bullying my man in the last one?

>>83847996
Holy sex. Can you do this with Millie & Enna, Laplus & Shion, or Sora & Marine?

>>83895258
These are great thank you for doing my request!

>>83826227
https://files.catbox.moe/6ltim8.png
https://files.catbox.moe/0blrvj.png

>>83881768
She won gold and will remain undefeated. Since breakdancing is not coming back to olympics...

>>83891989
Cute!
>>83895258
SEXO!
>>83896079
That is some sus looking Suipiss...

>>83891567

>>83762350


>>83891761
SuperMerge can do some 3 model merging but I don't exactly know what all the merges do, I just merge and make compare grids.
>>83895258
Oh painted on stuff, I've yet to try this prompt, you use a LoRA or need to innpaint? Or does pony models know this concept?
>>83896428
AI overlords wills Matsuri point at man with tiny dick


Good night /vtai/

>>83882668
not sure exactly what you want but heres something i guess
https://files.catbox.moe/qd3mbn.png
https://files.catbox.moe/07joq0.png
https://files.catbox.moe/mkqruk.png
https://files.catbox.moe/6bcna1.png
https://files.catbox.moe/e776v4.png
https://files.catbox.moe/ni2pit.png
https://files.catbox.moe/mnmt8u.png
https://files.catbox.moe/xg06jw.png

>>83899388
was able to sorta get it without a lora but made my own
most of my loras worth using: https://catbox.moe/c /6axzu1

painted clothes specifically:
https://catbox.moe/c /206j8p

should be baking a bunch more loras tomorrow if i dont end up fighting animated stuff for another day

>>83899388
Cute, Goodnight
>>83900409
SEXO!!

>>83891567
Me on the right

>>83819253
Does HLL-anon still visit this place? Can I ask a few baking-related questions, as a /hdg/ tourist trying to bring some artistic sense into Pony with vpred+ztsnr?

>>83823803
A bundle of Sara's saras coming up

https://files.catbox.moe/r2gyo4.png
https://files.catbox.moe/0iyotb.png
https://files.catbox.moe/pidfka.png
https://files.catbox.moe/xeddiw.png
https://files.catbox.moe/6wiflb.png
https://files.catbox.moe/8u92gy.png
https://files.catbox.moe/vh9qhn.png
https://files.catbox.moe/tz4ykw.png
https://files.catbox.moe/azhfcm.png

>>83885503
>>83886746
the comfy manager helps a lot with the transition from A1111 to comfy and there is a way to make your spaghetti straight lines, that help me to understood what was connecting to what more easily, is worth the shot, I'm starting to understand the limitations of A1111, but there are things in A1111 that I still miss, than I haven't found a substitute or a t least a proper one in comfy.

>>83840979
>>83841447
Wonderful Shiori's I like that you add the glowing eyes in a subtle way

>>83846720
Don't get discourage Anon, let me point you to things that actually help me to learn it https://openart.ai/workflows/academy this one is good to learn structure from basics to more advanced things.

>>83844855
>>83847165
Very cute style has a nice softness to it

>>83858493
Raden whole design fits perfectly to classical paintings this looks great.

>>83871284
Cute Selen, but the more you tell people that they don't do lewds more they will want to do them, specially if the design can help with it, who doesn't find hot a tomboyish girl with a ponytail, nice boobs and an eye patch.

>>83903591
It's usually polite to ask the question beforehand to speed up the process of getting a reply to it or so the other person can choose to ignore it

>>83903827
I'm pretty sure I've asked the question. What's the point in elaborating if HLL anon wouldn't even read it?

>>83903983
He's around every now and then

>>83903807
NTA but nice. Can I request a plump belly to match those big tits?

>>83899388
>https://files.catbox.moe/6wiflb.png
KINO thanks for fulfilling the request

>>83903807
Cute snek

>>83904644
Alright, thanks. In that case...
>>83903591
So I've been trying to train SDXL models to use v-prediction and so far I've kinda succeeded in converting ArtiWaifu https://civitai.com/models/615476 (kinda legacy) and Pony https://civitai.com/models/684052 (v0.1 is quite an old version at this point desu) using a really small, but hand-picked and hopefully diverse dataset, at around 2k pictures. Now I've expanded it to 5.5k with a goal of 10-30 pictures per artist which is currently in training (pic).
1. Why did you stick to using a lyco and not a full unet finetune? From my experiments so far, the "true" finetuning is the only viable option on a larger scale, while lyco tends to deepfry some random layers and freeze the others. Yes, I'm aware there's a "full" algo in lycoris, but it just doesn't work as well and suffers from these issues all the same.
2. Did you try freezing the text encoder? I've heard that if the text encoder is not completely broken, then there's not much point in finetuning it, and my experiments *kinda* align with that, although this would be hard to test on SDXL given I only have a 3090 and make test samples on a 3060.
3. Why you didn't really try baking for SDXL, specifically for Pony? Is it just the fact that it's too compute-intensive and/or you couldn't make it work, or was Animagine good for enough for you?

>>83906815
(by kame \(kamepan44231\):1.15)

>>83903591
>Does HLL-anon still visit this place?
Yeah, I check these threads regularly.
Feel free to ask

Also, If you train with kohya scripts, check if you are using options like min-snr-gamma and ip-noise. It probably won't work correctly with vpred and will cause problems

>>83907178
A lot of artists kinda don't work at all but here's also (by daizu \(melon-lemon\):1.3) with 28 images in the dataset, new epoch is in 20 minutes..

>>83903591
>>83906815
>Does HLL-anon still visit this place?
Yeah, I check these threads regularly

1. Why did you stick to using a lyco and not a full unet finetune?
It's smaller and uses less memory than finetuning. It made sense for sd1.5, because the size of lyco without conv was 1/3 of a full model. For SDXL it won't work that well i think.
>"true" finetuning is the only viable option
if you want to switch model to vpred. In your case you have no choice

> 2. Did you try freezing the text encoder?
If you train the text encoder, model will learn the names of characters and artists much faster . You can stop training te after training 5-10 epochs. Also you can try training only te2 for sdxl, or to use higher lr for te1 and lower lr for te2. It works better than same lr for both.

> 3. Why you didn't really try baking for SDXL
I'm not a big fan of pony(because of marine=aua bullshit). I trained some small loras, but i don't want to do anything big. It takes too much time

Random unsolicited advice:
- If you train with kohya scripts, check if you are using options like min-snr-gamma and ip-noise. It probably won't work correctly with vpred and will cause problems
- You'll save a lot of vram if you train in full bf16 with an optimizer that supports stochastic rounding or kahan summation.
Try StableAdamW, maybe. It's pretty good. Note that it needs slightly higher lr than usual
https://pytorch-optimizers.readthedocs.io/en/latest/optimizer/#pytorch_optimizer.StableAdamW

>>83819253
Hey I requested the tan Lumi pic from last thread I just wanted to say thanks for doing my request

>>83819294
Headpatting an exhausted aqutan resting on a couch after her graduation live, all while she has a smile on her face ;(.

>>83909431
Thanks for the detailed reply!
>if you want to switch model to vpred. In your case you have no choice
Initially, I thought that too. I've trained artiwaifutistic v5 via lycoris as well, and it just wasn't great. Much later I figured that you can merge lora layers to the models individually, so I tested that and I was really surprised that merging OUT5 at 0.5 gave MUCH better results. I'm not the only one who stumbled on this either https://github.com/kohya-ss/sd-scripts/discussions/294#discussioncomment-10081465.. This way I also figured that the only layers that matter for vpred conversion are out layers, to surprise of no one.
>If you train the text encoder, model will learn the names of characters and artists much faster .
I suppose this is true, but SD1.5 and NAI1 didn't train TE while they learned the characters still. I also don't want to train it for a few other reasons. First is that I don't want to ruin whatever was left after Pony with NL captions, since I train without them. The second is that my dataset is really small and it will ruin the TE no matter what I do. Does unfreezing really TE worth it, given I'd have to deal with those kinds of consequences?
>I'm not a big fan of pony(because of marine=aua bullshit)
I'm not a fan of that either, so I've been trying to unfuck it with zero success so far. "gawr gura" gives some abstract grey blobs (pitm still works) while "houshou marine" gives some faint red girl-shaped blobs. I believe it should be possible to teach Unet to also recognize those tokens, but I have no idea how to do that on a small scale except maybe something like a dreambooth, would like to hear ideas on that if there's any. That should be possible, unless Pony's TE completely ignores those tokens (which I doubt it does)
>It takes too much time
So true
>If you train with kohya scripts, check if you are using options like min-snr-gamma and ip-noise.
Kohya didn't work for SDXL vpred out of the box, so I had to fix it anyway (PR merged to dev branch). Regarding SNR, I use debiased estimation. It's mathematically correct for vpred as well and it works, but for vpred it jerks the loss like 10-15 times up so I had to lower the LR. Ip noise is also good.
>You'll save a lot of vram if you train in full bf16 with an optimizer that supports stochastic rounding
>Try StableAdamW, maybe.
Nice, will investigate that. I've been training with PagedLion8bit solely for the reason zerodiffusion and fluffyrock were trained with it, lol. But I guess with these pathetic batch sizes I use I should have used AdamW from the beginning.

>>83898596
This is exactly what I wanted, thank you very much anon!

>>83917250
Wonder if I should try this version of Lion https://github.com/warner-benjamin/optimi with fp8 base, kahan sum, fused backwards step, and full_bf16

>>83868417
>Would love to see more busty futa Holos from you!
Well I'd be happy to gen them for you!

https://files.catbox.moe/nv9iop.png
https://files.catbox.moe/34oznk.png
https://files.catbox.moe/eu9ohe.png
https://files.catbox.moe/iyty6u.png
https://files.catbox.moe/c2002c.png
https://files.catbox.moe/gu7s6w.png
https://files.catbox.moe/vvl4im.png
https://files.catbox.moe/w9a9bd.png
https://files.catbox.moe/2kquuq.png
https://files.catbox.moe/zqzswb.png
https://files.catbox.moe/1b2q56.png
https://files.catbox.moe/0zfw7x.png
https://files.catbox.moe/sfz77u.png
https://files.catbox.moe/x1xk59.png
https://files.catbox.moe/e1whh5.png
https://files.catbox.moe/0imik3.png

>>83896079
Absolute sexo

>>83847996

>>83896553
Guess who finally learned how to properly use both common and base for regional prompting? I feel like I've unlocked new possibilities...

No Laplus and Shion but I have the other two pairings; trying out adding more stuff.

https://files.catbox.moe/qdvsk2.png
https://files.catbox.moe/iykuoy.png

>>83848813
Thank you for sharing!
>>83860641
Oh yes, I have that, but keep forgetting to prompt using the LoRA.
>>83873545
Nice evolution of your style mix TWL, Bae's face in >https://files.catbox.moe/wyuxpl.png is really hot.
>>83891994
Beautiful - love the contrail and godrays
>>83900409
Is that a Loki lora?!
>>83903807
Nice Sara set, her model is sex.

>>83819294
Noefure, krofau, micomet or any other pairing holding hands and sharing a kiss while being fucked and creampied from behind. Any position.

>>83917250
>lycoris.. only layers that matter for vpred conversion are out layers
Maybe you can try training a lyco with --network_args "train_norm=True" (helps with stability) and a lycoris preset.toml like:
unet_target_module = []
unet_target_name = ["input_blocks.8.1", "middle_block.1", "output_blocks.0.1.transformer_blocks", "output_blocks.1.1.transformer_blocks", ...

It will train only blocks listed in unet_target_name. OUT5 is probably "output_blocks.4", but i'm not sure. You should check it.
Also add "LayerNorm" to text_encoder_target_module if you are training text encoder. That guy from a discussion page seems to recommend it too

>Does unfreezing really TE worth it
It's probably not worth it (if you want a very high quality model), but it can save you a lot of time
>I believe it should be possible to teach Unet to also recognize those tokens
It should be possible, but it could be very slow. If you train TE it may learn how to make a recognizable gura in (for example) 8 epochs, but if you train u-net only and have multiple characters/artists, it may need 40 epochs or more

>Ip noise is also good.
No? If I remember correctly, that's what was causing the problems in fluffyrock / EF.
Adding extra noise breaks zeroSNR. Model learns to expect and remove this extra noise. But there's no extra noise during inference so you get artifacts
Better to disable it

>>83919618
I didn't know optimi. Thanks for mentioning it. Looks like it supports stableadam too
https://optimi.benjaminwarner.dev/optimizers/stableadamw/

>>83917250
>lycoris.. only layers that matter for vpred conversion are out layers
Maybe you can try training a lyco with --network_args "train_norm=True" (helps with stability) and a lycoris preset.toml like:
unet_target_module = []
unet_target_name = ["input_blocks.8.1", "middle_block.1", "output_blocks.0.1.transformer_blocks", "output_blocks.1.1.transformer_blocks", ...

It will train only blocks listed in unet_target_name. OUT5 is probably "output_blocks.4", but i'm not sure. You should check it.
Also add "LayerNorm" to text_encoder_target_module if you are training text encoder. That guy from a discussion page seems to recommend it too

>Does unfreezing really TE worth it
It's probably not worth it (if you want a very high quality model), but it can save you a lot of time
>I believe it should be possible to teach Unet to also recognize those tokens
It should be possible, but it could be very slow. If you train TE it may learn how to make a recognizable gura in (for example) 8 epochs, but if you train u-net only and have multiple characters/artists, it may need 40 epochs or more

>Ip noise is also good.
No? If I remember correctly, that's what was causing the problems in fluffyrock / EF.
Adding extra noise breaks zeroSNR. Model learns to expect and remove this extra noise. But there's no extra noise during inference so you get artifacts
Better to disable it

>>83919618
I didn't know about optimi. Thanks for mentioning it. It looks great, and it supports stableadam too
https://optimi.benjaminwarner.dev/optimizers/stableadamw/

>>83920345
That Millie and Enna is great

>>83920345
I remember someone asking a similar request last year, and it was a nightmare to make. How hard was it for you now?

>>83920345
SEXO, Thanks! I like how intense Marine and daisenpai sora is.
Millie and Enna are also great. Their pussies are tantalizingly exposed, and I like the addition of a blowjob. It looks like an orgy might be starting.

>>83920345
SEXO, Thanks! I like how intense the sex with Marine and Sora is.
And Millie and Enna are also great. It looks like maybe an orgy is starting.

>>83921463
>(if you want a very high quality model)
That's the point, kinda.
>but it could be very slow
I suspect that this would not be slow at all once you give Unet a hint somehow. It already knows the characters, but old senile boomer pony's unet just doesn't understand alpha sigma slang. Maybe a dreambooth-style finetune for a few epochs is the key here after all.
>No?
>Adding extra noise breaks zeroSNR. Model learns to expect and remove this extra noise.
I mean, no - yes? It doesn't add the noise to anything but the "input" latents, that's what "I" is for? Why should that cause any issues with ZSNR? If anything, it should just learn to produce slightly noisy image every time, but due to the fact that the added noise also changes every time it just doesn't do that. And well, it learned to do solid black/white backgrounds perfectly fine (still struggles with magenta though).
>optimi
I think I might have heard of it on /lmg/ or /hdg/ once but this time I found it while searching for 8bit stable optimizers, not sure if it even works for sd-scripts

Which chuuba is your "muse" for when you want to try a new prompt or settings? It doesn't have to be your oshi per se. For me it's Mumei, I like her a lot and she's very consistent.

>>83923608
Ah yes, my favorite vtuber muse, the void.

>>83921463
>Maybe you can try training a lyco with --network_args "train_norm=True" (helps with stability) and a lycoris preset.toml like:
I did train it with "full" preset along with train_norm=True, but that's not the point. The point is that with current implementation LoRAs just update different layers inconsistently, so cheald estimated how fried each layer was and came up with alphas for each layer (instead of a global one)

furry https://files.catbox.moe/53uwff.png

Morning
>>83822470
Of course.
>>83923735
But that is my oshi.
>>83923856
>Kitsanya
SEEEEEEEX!

>>83923735
Take a guess

Is it weird that ai artists prefer Mio and Korone to all of EN?

https://www.deviantart.com/c18aiart/art/Ookami-Mio-Hololive-1034260456
https://www.deviantart.com/kw0337/art/Ookami-Mio-Hololive-989908934
https://www.deviantart.com/kw0337/art/Inugami-Korone-Hololive-989674866

>>83924622
Muh dick.

>>83924559
Kiara?

Dang her head magic circle thingy got a bit burnt in. Back to the drawing board kek

>>83923735
My go to is Fauna, she has distinct enough design, makes it easy to spot if something is working.
Cute moom btw
>>83923856
I am not a furry but...
>>83924281
Sexo! mornin
>>83924559
That's true we always need to do a frog test on all new models

>>83919780
Not the OR but holy sex these are amazing. Can you do some with Botan?

>>83920345

>>83920670
Using normal upscaling instead of hires - adetailer keeps fucking up the faces and fingers.

https://files.catbox.moe/4tql9o.png
https://files.catbox.moe/dpo6u1.png

>>83922266
Not sure, I only started genning about 2 months ago. Frankly these are just carried by the LoRA. I'm trying to revisit this FWMC-Shiori spitroast I did awhile back but it's still a bit gacha. Trying to also see if I can do a side-by-side doggy, but need more time to fiddle with prompts and regions.
>>83923735
Take a guess.
>>83925155
Thanks again for helping to bake this - looking forward to the final result. The halo is quite complex. How's the elbow gloves? Noticed they didn't come out in this gen.

>(skindentation:2)

>>83926008
it's like putting rubber bands on a melon full send it, go to skindentation:3

>>83926063
Captain, she can't hold out for much longer! She's about to burst!

>>83925155
Also trying to do this one, didn't like it initially but really grew on me after a while.

>>83925746
Holy fuck, based af gen.
>Elbow gloves
Oops forgot to to tag it in the gen, seems to be ok
https://files.catbox.moe/m7swxs.png

>>83926008
KEK, at some point the skindentation tag shifts to tourniquet tag
>>83926205
>Jumps for cover

>>83923608
> If anything, it should just learn to produce slightly noisy image every time
I worded it wrong. I meant exactly what you are saying. It will learn to always leave some noise, because inputs always have some noise. So instead of full noise at timestep 1000 -> fully denoised image at timestep 0 as it should be with zsnr, it will produce a noisy image even at timestep 0. It's like chopping off 100 timesteps from one side and stretching the rest. Technically, that's not a znsr schedule anymore?
It's fine now, but it will become an issue if training is very long. Zoom in and you'll see how this noise looked on ef-hll

>not sure if it even works for sd-scripts
it should just work if you install it and add --optimizer_type "optimi.Lion" --optimizer_args "weight_decay=0.01" ...
But it looks like things like gradient_release need some changes
https://github.com/kohya-ss/sd-scripts/pull/1381

>>83923782
I see. I only quickly skimmed though that that discussion on github. I should read it properly

>>83925292
Douzo.
https://files.catbox.moe/s9puh7.png
https://files.catbox.moe/f22vq7.png
https://files.catbox.moe/y85c23.png
https://files.catbox.moe/3jep4y.png
https://files.catbox.moe/sfvmsc.png

>>83926882
Incredible set, thank you! The only thing I could hope for is one with bush as well

>>83926974
Hopefully this much will do, Pony won't do jungles on its own and it's not something I'm into so I can't be bothered to dig for a lora.
https://files.catbox.moe/gnwk9j.png
https://files.catbox.moe/v6ajm4.png
https://files.catbox.moe/n0l2mt.png

>>83919780
>>83926882
>>83927371
SEXO!!! You are amazing, thank you so much anon. Forgive me for asking you to indulge me even further, but could you do all of Advent next?

>>83928992
SEXO

>>83927371
SEX thank you, it's not Botan without some bush

>>83837128
https://files.catbox.moe/n43ra0.png
https://files.catbox.moe/5lnwmv.png
https://files.catbox.moe/xiikw5.png
https://files.catbox.moe/2gjntg.png

>>83926547
>It will learn to always leave some noise
I see what you mean, but I'm still not sure about this whole thing hurting the model, since I'm able to produce fully black (#000000) images across all pixels. Since denoising happens in a latent space, a lot of things depend on how well this space is constructed, right? IIRC 1.5 VAE had issues with how it stored information in one blob somewhere or something like that https://www.reddit.com/r/StableDiffusion/comments/1ag5h5s/, so applying a "uniform" noise to a "non-uniform" latent space may neatly have resulted in the noise patterns you're seeing on your pic now, since those noise patterns basically represent VAE's compression ability. Since SDXL's VAE was trained properly, it should not have this issue really.
Regarding ZSNR noise schedule, IP noise adds noise sampled from a normal distribution, with noise values having both negative and positive sign. I'm not completely sure, but at 0 timestep given pure black image, shouldn't the model see an image that is randomly "too bright" or "too dark" (outside of 0-255 pixel range but still completely within the range of latent values), with an average being "perfect black"? In that case the ZSNR schedule doesn't break since only average values do matter.
>But it looks like things like gradient_release need some changes
Yeah that's what I was worried about

>>83932641
Also here's one funny thing I've thought of quite a long time ago regarding MinSNR and debiased estimation. Basically, all they do is rescaling the loss across timesteps, in order to make earlier timesteps receive less training on average. This means that some of the time the model almost doesn't learn, as at 1000 timestep the updates are extremely small, and they happen very often. So why rescale the loss if you can change the way the timesteps are sampled? Why not change the uniform probability density function that is currently used to sample timesteps to such a function that matches updated SNR amplification curve and leave the loss alone (constant)? That will surely lead to a more unstable training, but most of the training time will now be spent on the timestep range that matters the most, thus you can reduce the learning rate and compensate the instability. I *think* this shouldn't break the core diffusion principle because this approach doesn't really change the target for each timestep.

>>83929081
Requesting five at once is a bit much but I respect the thirst for cock.
https://files.catbox.moe/5grl2v.png
https://files.catbox.moe/5znbzw.png
https://files.catbox.moe/kbu7xk.png
https://files.catbox.moe/z5zsea.png
https://files.catbox.moe/dm3bdz.png
https://files.catbox.moe/llxl24.png
https://files.catbox.moe/cjn2a1.png
https://files.catbox.moe/0009g9.png
https://files.catbox.moe/sxh2yx.png
https://files.catbox.moe/of4gs9.png
https://files.catbox.moe/cu6ps4.png
https://files.catbox.moe/75sy32.png
https://files.catbox.moe/hv9enc.png
https://files.catbox.moe/yqbp02.png
https://files.catbox.moe/byhs27.png
https://files.catbox.moe/xgwwrm.png
https://files.catbox.moe/jayhxu.png
https://files.catbox.moe/uq4qsq.png
https://files.catbox.moe/x4yp31.png
https://files.catbox.moe/1kcqpi.png

>>83932345
Holy shit, very nice.
Thanks anon.

>>83867833
So much mayo.
https://files.catbox.moe/8jtwkv.png
https://files.catbox.moe/r7esrh.png
https://files.catbox.moe/bfb4ln.png
https://files.catbox.moe/jj100r.png
https://files.catbox.moe/svobzq.png
https://files.catbox.moe/vkgrnh.png
https://files.catbox.moe/dlb24u.png

>>83863401
>>83873545
>Nothing a little bit of photoshop can't solve
Very true. Not going to lie, looking at that amazing Flare Noel set again made me think about all the other chuuba combos. Also wondering if vaginal version would work too.

Coco receiving & Kanata cheering / Elira receiving & Selen cheering / Enna receiving & Millie cheering / Korone receiving & Okayu cheering / Kiara receiving & Reine cheering / Lamy receiving & Nene cheering / Mococo receiving & Fuwawa cheering / Mio receiving & Fubuki cheering.

It's very hard to pick one of these but if you're feeling like fulfilling this request, then please make Coco & Kanata. Also I definitely picture Coco being very smug in this scenario as in "See, I can take it!" but then having second doubts during the process itself. But ultimately the decision rests with you.

>>83876021
>but for NAI, on their wiki pages [...]
I see. That's pretty helpful.

>I had fun with the Asanagi style once
I actually haven't played with his style before. But yeah, as I said, check casino_(casinoep) artist. I'm sure you'll find his stuff to your liking.

>NAI has a feature that upscales your image and runs your prompts multiple times to "upgrade" the quality of the pic.
Oh, I think they added that recently? I didn't remember it when I had my nai sub.

>Since it upscales the gen, it costs tokens to use.
Doesn't it eat twice of your tokens to do so? I haven't used my tokens until my sub ran out so I had 10k extra to generate after the sub ran out.

>The higher the strength of the enhancing the more expensive it is, though I find such strength values unnecessary. I like using the minimum ones sometimes
So how much does it cost you normally with the settings you usually use, and what are they? I might want to try it out when I renew my nai sub.

>Makes the gen very smooth-looking and fixes a lot of imperfections.
I haven't seen the before/after so I can't tell if thats the case, you should know though.

>spoiler
So it costs you tokens for both upscaling AND enhancing, is that right?

>I do love to use the "cute" tag for everything so I can see that happening, lol.
Oh that's why. Makes sense.

>I'm very glad you enjoyed the set!
Thanks again for making it.

>>83891989
That's an awesome gen. Can't tell you what exactly it makes me feel, but, it makes me feel something.

>>83896079
How are you doing, Gooey-anon?

>>83920345
Enna is such a multitasker. Very nice gens.

>>83924622
Yuck. What an abhorrent style. I need some eye bleach after those links. Then again I should've known clicking these deviantart links.

>>83932641
>Regarding ZSNR noise schedule
Actually let me take that one back. Seems that it indeed does fuck up the ZSNR schedule in some way and makes it leave some noise, but it's pretty much only visible in img2img. On an earlier epoch this effect is much more subtle, so I guess I'll restart with that one. 27h GPU time spent well...