Happy new bread

>>23680 These gloves, very nice.

Anyone know what happens if you try to train a model on 768 or 1024 resolution but have images that are smaller than that resolution in the dataset?

>>23693 soft white gloves erotic

Insatiability requires teamwork

been struggling to do a proper emaciated-flabby kuroko all day instead of working

How do you merge the hll 6.3 file into easyfluff to defur it for future merges? Every time I try to use super merger for it it just says there's no lora.

>>23698 >it just says there's no lora. you have to let go anon the accident was a year ago

>>23694 nothing happens - bucketing

>>23698 supermerger is broken, you can open an issue ticket regarding that on the extension's github page

>>23698 >>23701 aren't you supposed to downgrade to a specific version to get it working?

>>23702 apparently, you just have to see which version

>>23704 Very cute

Brat in need of rape correction.

SDE+SMEA+sketch=sovl...

>>23708 yepp that's sovlful

Finally got a version of Based68 I'm satisfied with, too many days and I just decided that I wouldn't deal with vpred at the moment until there's a NAI vpred model or hll6 vpred model I can use to make it work better with NAI model merging. This version of Based68 was done by making schizo nai hll6 lyco merges with some of the models I put into the mixing recipe to make sure that some of the decent artist knowledge was still in there, it's not as good as hll6 and easyfluff but it still gets very notable artists right (you might need to up the tag of artist to 1.2 to get a more concrete appearance of their style) https://files.catbox.moe/sizp9i.txt <- txt file filled with the artists https://gofile.io/d/nKxELQ <- model will do a public release after some personal and anon testing

>>23710 Neat, I'll check it out soon.

>>23710 Where does the txt file go and is it needed? I assume as much since you've included.

>>23712 oh the text file is just an artist wild card extension, if you have that extension you just add it to that otherwise you can easily pull up the artist with the autotagger with this .csv file, you just have to add this to the autotag complete extension tag folder https://pixeldrain.com/u/HRpYWizR

>>23710 but artist tags work well and even without it, LORAs will still look good without having to use artist tags that are embedded in NAI hll6 fluff

>>23713 From my use case I'm getting some garbage but it's 100% user stuff, messed with cfg rescale and it cleans up the image a bit. Got a catbox or some tips for usage with this since it obviously handles differently then previous iterations.

>>23715 this one doesn't need a specific yaml or CFG rescale, it's still a base 1.x NAI merge

>>23716 Guess I'll try restarting it and see what it does. I've tried with loras, without and it's pretty fuzzy. Might just be an issue loading it since I just downloaded it and refreshed my list instead of restarting sd.

>>23717 I have to make sure on my part too, I'll restart because sometimes automatic111 will keep the supermerged model on even after saving it

>>23718 >Supermerger Yeah I was messing with it so that might be fucking me. I'll say though b68 seems to blend really well into other mixes I was messing with.

>>23719 the meta data is complicated because a lot of the merging recipe is using Easyanon's old lyco merge.bat script and just merging hll6 nai into the models I wanted to use to make sure all the merges still kept some of Hll6 nai's artist knowledge. It's a lot more weaker compared to the Easyfluff version so that's the downside of it but it looks really bad with NAI based models so meh, nothing much I could have done there.

>>23720 I'll fiddle some more with it since one of hte earlier b67's was my favorite of yours by a long mile since b64. Guess I'll bite the bullet and update to 1.7 too and just do a full upgrade, see how bricked I can go.

>>23721 Ganbare, looking forward to the stuff you anons share, gens and opinions I'll keep a close eye before I feel chill enough to release it outside of here.

some lazy river prompt testing,

>>23721 Ooh that b67 looks very good, is it the same that si uploaded on civitai? If not, I would like to have it please

Have local prompters all roped yet?

have you stopped being brown yet

have you stopped being a nigger yet

sorry double-post

in unrelated news i am once again banned from 4chan for racism. does anyone have any suggestions to avoid this in the future? being less racist is not an option.

>>23729 Stop being triggered by crap. The more You fuck around the more You find out

speaking of merging.bat scripts, easyanon if you're still here and working on the GUI is it possible that you'll be adding the lora/lyco merging option to it in the future?

>>23730 i practice a gentle, paternalistic form of racism. someone has to periodically remind non-westerners that the reason their countries suck is because they're full of people like them, and perhaps if they practiced being less like themselves their countries would improve.

>>23732 That one is local? Catbox anyways pls Also It's not like one hundred million copies of yourself would be better for your country as well... I mean... One hundred million weak and petty losers who jerks off to AI anime porn would be really bad

>>23733 the info's all right there, projecting brownie-kun, but here you go anyway: https://files.catbox.moe/9498xh.png

>>23732 >and perhaps if they practiced being less like themselves their countries would improve you're not wrong and those who post on 4cucks probably already know that anyway t. thirdie

I haven't been around for weeks, I don't understand why the sudden hate towards people who don't wanna use corporate controlled models tied to subscriptions.

>>23736 Neither do I but I also won't hate on people that are paying for NAIv3 either.

>>23736 >sudden hate it's just mutual trolling and a bit of schizophrenia.

>>23736 imagine being this fucking retarded. get off the internet if your IQ is too low to tell trolls apart

>>23729 you sound like a nigger and im not talking about race

>>23740 it was maybe funny first 2 times he posted it but now this is some "how do you do fellow racists" cringe

Based68 is out for everyone, I don't know why but it doesn't look as sharp as 67, it kinda has a Based64 vibe... well maybe my "Based64/7" model I merged is the reason for that.

>>23733 That doesn't look like local, don't give me hopes.

>>23744 >prompt guidance rescale 0.7 why touch this?

>>23745 Depending on prompt(especially artists) and cfg, without rescale the images can look kinda oversaturated, almost fried. 0.7 was probably too much for those though.

Is hydrus import script anon here? I found two cases where the script crashes. One at line 567, when the webui meta doesn't have a negative, because negatives = None at line 460, which I fixed by changing it to negatives = "". Second is for nai at line 635, when the meta doesn't have "Generation time". This one is more complex for me to figure out, see pic.

Be fruitful and multiply, and fill the earth and subdue it; and have dominion over the fish of the sea and over the birds of the air and over every living thing that moves upon the earth

>>23748 need more shondo

>>23749 based

>>23747 Also noticed it's very slow compared to the old june version when checking images that have no data, like seconds vs hours slow.

Fuck, the enhance gatcha gets me. I can't choose which one I like best, some I like the face best, some have better butts...

>>23753 probably the last one

What's the best Dagasi lora out at the moment and the best checkpoint to use with it?

>>23753 I love vampy lewds Wish she got more art

>>23756 Be the change you want to see in the world.

>>23758 my wife chino

>>23758 Very cute

>>23760 Catbox for all of those please, localbro

>>23761 Sure thing but I'm actually using both naiv3 and local now. https://files.catbox.moe/aiwbn6.png https://files.catbox.moe/ujzspc.png https://files.catbox.moe/1dep43.png https://files.catbox.moe/usmq5b.png

Slightly plump

Colonizing cunny.

>>23765 >Born too late to explore the world and colonize cunny FUCK

>>23767 Mana transfer erotic

>>23767 had annoying fuckup in one of the pics >>23768 certainly

>>23749 based fishfucker ruto > garbage > mipha

>>23764 (healthy:1.5) >>23765 Thanks for the inspiration.

Measure your shit right or you too could end up with a tiny bathtub.

>>23772 based bathing enjoyer

>>23772 >>23773 One week and one tape measure later

I made way too may of these

>>23771 How do you get them this brown without any big emphasis anyway? I checked a bunch of your gens and you don't seem to have anything out of the ordinary. Is it the zankuro lora? I had to do like (dark skin, dark-skinned female:1.5) and they still turned out pretty light skinned.

>>23776 Zankuro probably helps but I'm also using the kasshokuhada LoRA pretty often now, it gives them darker skin.

why did i check spoilers this css is confusing

My schizo daughterwife

>>23778 Nice. I should have done more bathing gens...

>>23778 >>23779 Nice.

Mleiry Christinas everyone

>>23783 Mleiry Christinas

>Rediscovered controlnet >Have to stop myself so I don't spend all xmas fiddling with dumb shit Merry xmas bros

https://huggingface.co/JosefJilek/loliDiffusion/tree/main The loli diffusion guy seems to be up to v14 now and taught me about loli models I didn't know about like Rabbit https://civitai.com/models/121696/rabbit

xmas merry

Merry Christmas

Merry Christmas

>>23789 >>23790 Merry Christmas

anybody have the haguhagu_(rinjuu_circus) lora?

>>23792 https://files.catbox.moe/598sod.safetensors

>>23793 Thanks!

>>23793 you could land a plane on that 6head

Based68 version 2, no major updates just updated some of the models I used in the first version with the updated version. HLL anon has released a lot more hll6 models but they're all fluff compatible so it won't look as good with NAI models, so I'm still using the NAI version of HLL6 with the merges. https://huggingface.co/AnonymousM/Based-mixes/blob/main/Based68-V2.safetensors https://huggingface.co/AnonymousM/Based-mixes/blob/main/Based68-V2-Pruned.safetensors

wow Civitai really are Australian politicians when it comes to sfw/r-15 cunny, rumao well at least the huggingface link still exists

>>23796 Cute and hot. >>23797 Neat.

Welcomed home

It's fucking over fizrotbros https://baraag.net/@fizrot/111655105584621263

>>23801 F

Regional prompter is very strong even for prompting concepts that the model isn't trained well for. That said, gawd damn it's a lot of work to get this one. NAIbros, how hard is it to make this pose for you?

>>23803 You can get it with the right tags, extra weight on cunnilingus and anilingus tags, plus some luck. Would be much easier to inpaint but the failed gens are also really good

>>23804 Every time right after I post

>>23805 Cool, thanks for sharing! I'm getting decent results now but the promptwork is a bit nuts. Using controlnet on a good gen is a massive help

Never tried facesitting with Naiv1 based mixes but it works pretty well with Naiv3.

>>23807 Update, I think I got the regional prompting required for this pose down >>23808 With strenuous regional prompting it could probably be done. Haven't really tried yet though

>>23809 I never thought about using regional prompter in that way, maybe I'll give it a go at some point.

K the last ones for now, gonna try something else.

>>23811 Third one is pretty nice.

>>23797 is there a specific way to use this model? because the images im getting are fried.

>>23814 one comment said it had a orange tint but nothing like that ruamo, don't use CFG rescale

>>23813 I'm liking the soft shading pseudo-3d that nai does

Potato wives best wives.

I'm addicted to prompting bestiality in fluff. It outputs really good quality, way better than any nai merge combined with all the bestiality loras. Human eyes on dogs are really annoying though. No one bothers to tag them properly, so it's near impossible to remove them with negative prompt. Nai merges, on the other hand, always produce realistically looking animals, but they output nothing but eldritch horrors when you ask for sex. The only one that's sort of good at it is the fresh Based68.

>>23818 plump

>>23818 you should do: mahira (granblue fantasy), {{{{{{fat mons}}}}}}

>>23819 I was pretty disappointed with the results when I tried to get it to gen human on feral female. What do your prompts look like?

B68 v2 pretty nice but ef still does lolis better

>>23824 Cute, I have yet to give B68v2 a try but I'll do that next year.

>>23824 which is which? right does child bodies better but I prefer left as im not sick in the head/browsing TOR at the moment.

>>23824 I like left more.

>>23822 It's usually a prompt consisting of 3 parts: 1) general human language description 2) styles 3) refining tags Example: [human female penetrated by a huge realistically depicted domestic dog animal from behind, realistic, [werewolf|dog], bestiality, perspective, fisheye lens, ::0.5] END (by butterchalk:1.2), (by pumpkinspicelatte),by smewed, END (human:1.1), (domestic dog:1.1), bestiality, female on feral, female,feral,penetration, knot, younger female, adult on young, anal, anal penetration, pussy, long hair, tears, balls, pussy juice, genital fluids, huge breasts, sagging breasts, all fours, When I need some unique artist, I add a lora (even nai-based) to the end of the prompt (3) and add the artist tag to (2) or use hll lora. It's gacha. After generating something I like, sometimes hires.fix does the job, sometimes it's fatster to switch to nai-based merge for img2img.

Happy new year

Just here to say Happy new year y'all!

>>23830 >>23829 Happy new year!

>>23811 >>23817 stealing these, I'm a sucker for that fizrot-esque pseudo-3d

>>23811 > 3 Please tell me how you did it. It's amazing.

>>23833 Use the catbox extension, it's NAI >rurudo lion, {{artist:dino (dinoartforame)}}, artist:s16xue, [[artist:fizrotart]], [[[[artist:dorontabi]]]], solo, 1girl, standing, ass focus, bent over, leaning forward, arched back, elbow rest, looking back, :o, skin fang, lion ears, extra ears, lion tail, nude, cleft of venus, hair intakes, ahoge, long hair, yellow eyes, indoors

>>23834 >{{artist:dino (dinoartforame)}}, artist:s16xue, [[artist:fizrotart]], [[[[artist:dorontabi]]]] Unfortunately I'm getting pretty fried eyes with that, gonna have to import one of the images and copy the other settings

>>23835 Maybe try enabling smea (and DPM++ SDE) or add @_@ in negs

>>23836 >Maybe try enabling smea (and DPM++ SDE) or add @_@ in negs Done, done and done.

Been playing around with B68v2 for a bit now and yeah it's pretty good, gonna give that furshit a try as well.

>>23839 cute

https://arch.b4k.co/v/thread/662498484/#662569712 heeeee

>>23841 I'm more impressed that 4chan wasn't already pre-blocked. If true, anon might be in the clear because their IT security sucks ass >t. ITfag

>>23841 yeah he sure would be typing that out if thats what was going on. dogshit bait fanfic

>>23841 reading through /v/ threads feels almost same to reading through twatter

i have a problem and keep downloading full checkpoints instead of style loras

based68 is based i tell you hwhat

>>23847 Nice. >>23848 >>23849 It is but I've been giving easyfluff with hll a try now, I have sinned... and that Mikan is perfect.

Is based68 one of those vpred models? I still don't know what they really are and what makes them special. Can I get a qrd?

>>23850 i tried it too but didn't like it. i prefer the vibrant, sharp and flatter shaded look of based68. pic is fluff with similar workflow and artists. maybe a skill issue.

>>23852 I apply the color autofix with the CFG Rescale or whatever it is called after I have done all my inpainting and stuff. I just img2img with 0 denoise and tick that option.

>>23853 But I would pick B68v2 any day over fluff for 1girls though.

>>23851 Based68 is just a merge between Based64 and the Hll6.3 lyco LoRA. It does not use vpred at all.

>>23855 So the only thing it has over regular models is artist knowledge?

>>23856 it seems to be better at concepts and poses compared to other anime models from my experience. also with artist knowledge comes character knowledge so you don't necessarily need a seperate lora for some characters.

>>23857 I see. So I don't need anything extra to make it work and old loras also work?

>>23843 he wrote those words to make you smile and you chose to be mad and gay instead

Somebody stop me.

>>23860 I stopped myself, one of my fan's ball bearings seems to have decided to shit itself.

>>23858 Basically, yeah. Might want to have in hand the artist's names that the model (or the merge lyco) recognizes

I'm gonna start collecting artist datasets now that I got some more free time and file storage Any artists (preferable if they have lots of stuff not on boorus unless its past the NAI model cutoffs) that people would wanna see?

>>23850 Cute cows, hope the pc is in working order again soon

>>23864 Well the fan is spinning now but making a lot of noise, I still have my warranty so I'm just gonna take it to the local store I bought it at.

Hey all, I haven't been around for a long while, been pretty busy with life, but I was able to work on the UI a bit this new years! It's only on dev right now, but I added support for lycoris's new OFT model type among a ton of other bug fixes. I also updated (remade) the installer so that it better supports the newer torch versions, so now anybody who installs my scripts will be using torch 2.1.2! A few small bug fixes have also been made, and I plan on pushing to main either later today or tomorrow. The next thing on my list is to get a new training page for full scale fine-tuning up and running, 1, and get the lora utils added, 2. >>23731 As for this, yes! I do still lurk occasionally, but am unfortunately not able to be around as often as I'd like. I do have plans for adding those, but you'll have to wait, as my time for devving the project has been cut into a 10th of what it was.

>>23866 Oh, and, I'm not sure if anybody noticed or cares, but I also released all of the character lora that I've created over the last few months onto civitai a little bit ago, check them out if you want: https://civitai.com/user/derriandistro/models

>>23863 U mean for what model?

me in the back

Here I go training loras again

>>23871 Nice, I should test how good this model is at bathtubs/bathing and that goes for b68v2 as well.

>>23872 Wonder if the bathtub stuff will work. Are you prompting using the e621 tags, the booru tags or a mix?

>>23873 A mix of both, I'm currently working on something else but I'll give the bathtub a try when I'm done.

>>23874 It has a decent understanding out of the box imo

UOHG, toasted cunny is best cunny

>>23875 Yeah that's good, I can work with this for sure.

Well this sure was easy to gen with Easyfluff+hll

>>23859 What kind of artist combo gives this

>>23872 based68 does tubs consistently but it also wants the whole tub to fit into the image lol

>>23880 I was gonna try both but coolio.

>>23879 oh right doing a composite in photoshop obliterates the metadata iirc it was just fizrot with a little bit of murakami suigun for nicer bondage

>>23849 Did you inpaint faces on the third one? The left looks like ishikei, the middle is ke-ta and idk which is the right one >>23871 Catbox for 2?

any models to recommend? i got based68 but no matter the version i use i get fried images

>>23883 Sure >>23884 For me it was an issue with my vae. Afterwards I got based68 to work just fine. Easyfluff with hll6 is also pretty good. For that one you'll need some extra stuff cause it's a vpred model and you can use mix of e621 and booru tags

>>23883 no inpainting for that one. i genned a base 3girl image for cn lineart and use that to align regional prompter with different artist tags. 3rd one is mignon. running the "final" image through img2img with cn tile at lowish control weight and lowish denoise strength can refine the image without necessarily needing to use masked inpainting. seems to cost some saturation, though.

>>23886 Based and hot.

Yes, I needed multiple versions of that image.

Succubi are also hot

>>23890 That's true and I don't gen that enough.

>>23885 Thanks, anyway, got some experimental versions of that 40hara lora, if you would like to test the FR or nai version, probably overkill by now, since ef+hll is already can do that style pretty well, at least on a7 version >>23887 Interesting workflow, but did i remember incorrectly or regional prompter wasn't working with hires.fix part? > running the "final" image through img2img with cn tile at lowish control weight and lowish denoise strength can refine the image without necessarily needing to use masked inpainting. The refined image looks pretty nice, i'm interested about what values were you using for that, could you push a catbox example, so i could figure out the technical part of this myself?

>>23892 it's gone. https://files.catbox.moe/p5wog9.png i'm generally using 0.25 control weight and 0.5 ending step without a preprocessor. i haven't noticed anything wonky with regional prompter with hires. fix.

>>23893 I think i confused regional prompter with another extension with similar functionality. Trying with preprocessor and without gives me the same image, dunno why, but overall refining is pretty nice still. > Processor Res: 64, Threshold A: 64, Threshold B: 64 Cant find this settings in the cn menu, is this something important?

>>23894 nah, i only changed weight and end step

Anyone know if there's a lora of this princess connect character? I found one on pixai but pixai sucks fucking ass and downloads are disabled anyway

>>23892 No worries. I'm always down to test some versions of 40hara loras. Ef+hll can do the style pretty well, but it doesn't know characters like Chitose automatically. I still mix in the 40hara lora from civit to get the maid outfit correct

>>23892 Forgot to say, nice chitose. Also, were can one find an updated list of the hll supported artists? I know the new version can be found here https://huggingface.co/CluelessC/hll-test , but I can't easily find a list of artists

Can I somehow extract this hidden information of yours from already downloaded images? Dear creator of the catbox script, can you make another one so that i can run it in a folder with images and output the result to a file, please?

>>23900 https://warosu.org/vt/thread/65918142#p65956121 https://warosu.org/vt/thread/65918142#p65958315 Direct links: artists - https://litter.catbox.moe/a1wmhj.txt tags autocomplete - https://litter.catbox.moe/q5w3i4.csv wildcard artist list without the #digits - https://files.catbox.moe/nncv7p.txt

The Kirin armor LoRA works with ef+hll, it's so fucking over for me.

Just got back into proompting after taking a break for half a year. What's the new state of the art for local models. I see mentions of ef+hll here, what is that? Has there been any new tricks to Lora/Locon training?

>>23904 ef+hll is Easyfluff + that hololive/vtuber lyco/lora. You use hll to defurry Easyfluff, it also let's you prompt artist styles that's in hll by using "by artist". I will be using ef+hll for actual sex and b68v2 for 1girls. Here's a guide on how to get started, ignore the comfyui shilling: https://rentry.org/5exa3

>>23904 What did you use for Fu xuan and Hanabi? Can yo share if possible? >>23905 Dude it's insane how much comfy shilling there is these days. The one AI site I use for free genning these days added a workflow mode and it's literally just a comfyUI interface.

>>23906 Yeah and the more shilling I see the less likely I am to ever giving it a try.

>>23907 as a free user, I can't really be picky since I don't have my own GPU to gen on I haven't messed around with the workflow mode but it's probably the closest i'll get to local genning now that colab is fucking dead It's just such a shame comfy is so much worse than A111

>>23899 >>23900 > Forgot to say, nice chitose. Ty, sorry for late response, here the FR version of 40hara which not that great actually, but anyway https://files.catbox.moe/wit1jh.safetensors And this one is i think the best shot of all the tries for 40hara dataset, was trained on nai, but should work with ef+hll still https://files.catbox.moe/7suthr.safetensors Since links for tags and artists from another anon are dead already i'll reupload them too https://files.catbox.moe/58105m.csv https://files.catbox.moe/cmg3nw.txt

Isn't there a map tagging that basically has the Tag equivalences from danbooru to e621? Cuz the e621 tagger is not that great actually

>>23896 Can I request for someone that actually knows loras to try making a lora of this character? I made an attempt but it's basically only usable at 0.3 since the picture and the quality of it get fucked if I go higher Here's the images used https://imgur.com/a/9t5eG6M https://danbooru.donmai.us/posts?tags=quria_%28princess_connect%21%29+&z=5

why does 4ch hdg hate local now

>>23912 like 60% of the shitposting in any recurring thread comes from the same small group of people running from one end to the other like they're trying to rock a boat over

>>23913 It's just so dumb I miss the /hdg/ that felt more like bros helping each other and finding ways to improve their gens over the >Just buy a NAI sub, are you poor or something??

>>23914 >he thinks it's about owning poorfags and not a quality/effort ratio do localoids really?

>>23916 pretty good inas

>>23914 everything good is always fading

>>23905 Thanks bro, I'll check it out >>23906 My homebrew Loras. They're not the best quality. I had to do a lot of inpainting on the Hanabi image in particular. Here's a few more

>>23919 Hanabi looks good, I'm surprised you actually managed to get the mask right since sd always hates on masks.

>>23909 Thank you anon, I'll give them a shot soon >>23914 I think plenty of us are still here to just share gens and help each other improve. Ignore the bait, stay positive, share some good gens

>>23912 Most of it is just a stupid bait but retards with "lol nai actually sucks just inpaint bro are you a shill??" are genuinely annoying.

>>23911 If your fine with normal nai based models I'm fine with giving it a try

>>23923 Sounds great, please do.

>>23893 Is there really a point to that if you're not upscaling or style transferring? Or do you mean you do this in multiple steps little by little after hires fix, so it adjusts to the new res better in less drastic way?

>>23925 it just works. even if you use cn tile and high denoise for the final upscale, passing that output through img2img again at same res, random seed, low control weight and low denoising strength, it gets improved. imo. if i used hires. fix, it would go something like this: >hires. fix to 1280x1024 >upscale with img2img to 1600x1280 with or without cn tile >pass it through img2img with low denoise and low cn weight >maybe pass it through img2img with low denoise and low cn weight again >inpaint final touches

>>23869 what model? sir...

>>23927 it's nai

>>23926 Great stuff, that pussy peek is really nice.

>>23928 It's over...

>>23930 not if you have 25 bucks /s

>>23930 sorry dude

>>23924 It's done, took a bit because I was at work, and the number of images meant testing was more important than usual, works fairly well, can do nudity, didn't test much more than just being nude though https://civitai.com/models/258865

>>23933 some more nudity, sfw examples are on the civit page

>>23934 >>23933 wow this looks amazing, thank you so much.

>>23935 no problem! The fact that you gave it a try yourself was the main reason why I wanted to try and do something nice for somebody. (that and obviously, more cunny is good) Either way, hope you enjoy using it. It's not often that I make characters with only like 20 or so images, so it was a fun change of pace for me.

Wtf NAI why you pasting a dick in my completely non-lewd image

>>23708 Has anyone been able to get Wagashi's style working well? I've tried but it always seems to mess up faces a little

>>23939 shit i didn't mean to reply to that

>>23931 the audacity of genning anime cunny on something tied to your credit card...

>>23941 There are ways to pay indirectly (for additional fees). They really aren't gonna blackmail you or call anyway but you can take basic precautions if you are feeling paranoid. I love the cunny it makes so I just can't resist.

>>23941 >the audacity of genning anime cunny on something tied to your credit card... what are they gonna do, arrest me? lol imagine living in the cucknited states of jewmerica

>>23943 I kneel.

>>23941 >>23941 >>23943 third worlders inventing reasons to compliment themselves again

>>23942 How do you pay for NAI without a cc?

Looking at the openmodeldb site (apparently where upscalers wiki moved) and it seems like there's quite a few new models. I've been using lollypop for regular upscaling for as long as I can remember, are any new ones much better?

>>23946 visa gift cards work

Found this model hat seems kinda nice https://tensor.art/models/632563024143657288 The author disabled downloads but I managed to grab V2 while it was available here if anyone that knows more about models wants to check it out since idk if it's actually good or not, I just gen casually. https://mega.nz/file/C7QUwYKD#OQmc1ZUU4vp7DYAELaTMAhe-jxGVmIj5E5FVOT-0Wdg

Anyone know how to set up Sagemaker without having to bind it to my real person? I want to set it up but I don't want to give them my full data

hey onii-chan check this out

heres your local sdxl anime model https://civitai.com/models/257749/pony-diffusion-v6-xl

>>23955 heard it's very good but the creator is a bit schizo, especially for style he doesn't want artist tags at all and prefers descriptions of the style instead

>>23893 >"PuffyPussyv2: 8cec737fcb1c" Where can I find this? Can you share please?

>>23955 Spat out rainbows for a while, but after I changed the vae to something sdxl compatible, it's workable. Seems to be pretty good at anatomy and composition, but the style is all over the place if you don't specify anything, and figuring out exactly what to describe feels like it's going to be a pain. Also takes a good while to generate on my outdated gpu, but I'm pleasantly surprised I can run it in the first place. And I noticed my colors looked washed out. Does it need the cfg rescale too, like the furfag model?

>>23958 it's not really needed and you can usually get away with just (pussy, cleft of venus, innie pussy:1.1) or something like that and have clitoris, spread pussy in the negs. based68 seemed to struggle somewhat for whatever reason. https://files.catbox.moe/8gjfim.safetensors first one is without, second with and last is only the final upscale with it (i do this). i tried to make it as light as possible but you can still see some style and eye change. i didn't crop the heads out from all images so maybe i'll try that.

>>23945 mald

Why do the furries get all the nice free models

>>23962 because they're the ones doing the work

>>23963 Animagine 3 and Pony Diffusion v6 just came out and both are intentionally gimped by pruning artist tags

>>23955 >Any explicit content involving underage characters has been filtered out. Useless garbage

>>23955 Only the two-faced furrfags are able to train a model full of the most vomit-inducing zoophilia, while also taking the moral high ground by purging it of cunny.

Animagine could be good if it's ever finished

>>23966 That's the thing with furries, they have MOTIVATION and money but they're way more schizo than us. Meanwhile Anime fans don't have the money with all the autism they contain

Does NAIv3 know Masuda's style? I noticed his page was taken down from danbooru

>>23969 doesn't seem to

>>23966 I've seen people denounce those who like anime cunny as literally hitler then turn around to masturbate to guro and yaoi shota so im not surprised anymore by these things

fiz-rot deleted his pixiv, misskey, and baraag pour one out for a real one

>>23967 >>23964 Do you gen on sdxl locally or do you use some service for it?

>>23972 it's just a way to feel better about yourself >Well yeah I masturbate to gay guro knot scat vore, but at least I'm not like THOSE disgusting cunnyfags!

>>23965 Ah, no wonder I felt like I was having trouble getting flat chests. Into the trash it goes.

>>23972 funnily enough cunny is the only normal one out of them all. at least assuming its not single digits.

>>23974 Local, SDXL works fine in A1111. If anything I get less vram errors genning at high res with SDXL than SD1.5