Im really enjoying the Pony based models. So much better prompt control and less hand fuck ups. And It only took one try to get a good looking blob.

>>24007 Excellent sequence!

>>24007 What's this from and how did you generate

Here are some generations I made yesterday. Could anyone help me figure out why Stable Diffusion isn't making the belly any bigger or Susan any fatter?

>>24019 Thanks, those are pretty good ass shots. Are those cropped or were you able to get the ai to focus solely there? >>24021 These are all just solo,1girl stuff I'm messing around with. I used the T-Ponynai3 checkpoint on Comfyui for both of these. >>24028 You have your lora strength set extremely low. I can get Symix to work just fine at 1.00, in fact, these are all using Sy-Pony at strength 1.00. Also, the AI likes to try to keep everything you prompt framed within the shot. Since you want wide women, it helps to set the latent to a landscape resolution so it has more space to work with.

>>24029 > Are those cropped or were you able to get the ai to focus solely there? Not cropped, just prompted with "head out of frame, from behind". I have a few from the front too. >>24028 As the other anon said, 0.15 strength is way too low for symix.

>>24085 Could you catbox the original unedited version please? Ive never gotten navel fucking before.

any prompt suggestions on making a fat belly more prominent and large on playground? I'm not getting good stuff with "large fat protruding prominent flabby belly. Soft obese pale doughy gut with muffin top and large love handles."

I'm new to using XL models so I'm trying to get used to it with Pony XL, but here's two alright ones.

>>24258 How'd you get clean results with an SDXL model? I'm just getting all sorts of muddy results like these

>>24265 You're gonna have to post your gen settings and prompt or else we'll just be guessing at what you're doing and not doing.

>>24311 Not how this works dumbass.

I'm kinda figuring this out.

Reference sheets are pretty fun

Naoto

>>24513 nadine the worst ome

From a prompt on /d/ today > a cute /d/ girl confessing to the viewer

>>24323 Please provide the Prompt used and negative prompt too.

>>24510 yesssss that's so fucking hot, more fattening zelda plz

>>24541 Here are the catboxes: https://files.catbox.moe/wg4mo9.png https://files.catbox.moe/p0glsf.png https://files.catbox.moe/bcf4dk.png

>>24510 Holy SHIT these are good, do you think you'll do other botw characters like Mipha and Urbosa as well?

Hey all. You might have seen that I've recently made the switch from SD1.5/Aurora Borealis to PonyXL/Symix. Today, I was getting a tad bored of just the handful of artists within Symix, and was missing the variety we had with all the custom single-artist loras that were made for SD 1.5 (Trinity Fate, etc) that never made the leap to SDXL or Pony. So, I started falling down a rabbit hole with style loras on CivitAI. I tried 100 of the best rated or most interesting looking style loras, with the same fairly generic character (sae chabashira) with the only thing really changing between loras being the seed and what expression. If you're like me and just crave variety but don't exactly know what style might be a good fit, have a look here: https://gofile.io/d/QflVfC (if you want them all zipped up for viewing on your own time/device or for when this link eventually dies: https://gofile.io/d/S7M4O4) These were all generated with a combination of the character, Symix, add detail, and artist lora, so 4 in total (probably the upper limit for coherence). Just select 'play' next to any image you want to view (you don't have to download), the name of the file should lead you to the Lora on CivitAI if you're interested in trying any of them for yourself. These are all un-edited as well so you can check the metadata for other information if you wish. Hope this helps someone! As for me, the top 5 "new to me" artists would be biggydeez, Limebreaker (very BWS-ish), Rapscallion, Sakananoike, and Valbun. Oh also, quick note, these were all run on source_anime, and some probably could have fared better on source_cartoon, so don't judge some of the more fugly ones too harshly, they might look better with some tweaking. Same for some of the faces, they all had a generic adetailer pass that helped most and hurt a few so don't judge the LoRa in that regard either. I'm sure many of these could be improved with some small changes. This is more to give you a feel for the general composition and fat compatibility of these styles.

>>24565 In case you weren't aware, the artist LoRAs are also trained by me, the maker of Symix. They were prototypes for Symix itself. As to why all of the artists trained as standalones aren't in Symix? It's extremely labour intensive to manually review auto-generated tags, tweak them and add custom tags. They will make it in eventually, when I find time to add in more of them. Good on you for trying all these different styles to add more variety to gens! I'll be taking a look at that style grid.

>>24566 That's awesome you're still hanging around! Thanks so much for all you've done making so much of what we do possible. As cool as multi-style and multi-character loras are, I can see why some creators choose to separate them rather than bundle more and more together. On the one hand it's super convenient to have one less lora to load (my general fat AND artist style in one lora, heck yeah!), but at the same time the individual artist loras added their own utility. Yea you know, I was kind of in my own little bubble with Symix, just using the artists in it only and sometimes combining two of them, but never more than that. So just on a whim I started trying style after style just to see what was "compatible" with my current workflow, continuing to use Symix but not specifying one of it's built-in artists. Most ran just fine at 0.8 like everything else, some needed a 0.7 or a full 1 to not glitch out. And man, yes, there are a lot of generic "anime waifu" type artists, but at the same time, there are some really interesting ones out there I've never heard of or seen. Many of course are wishy-washy on fat-compatibility (have to push some tags really hard to get anywhere) but it was a fun experience to see all these different types of fat depictions when I had grown so very accustomed to just what the current version of Symix contains (which is still a great selection especially if you mix and match).

>>24568 Yeah I try to keep an eye on most places where my LoRAs are used. Even if I kept the artists separate, I would still need to update the tags to have the size control tags (belly overhang, belly on floor, double belly, double chin, that kind of thing), so as to maintain my standard of quality and controllability. Those artist LoRAs might be good enough for many people but they lack the control and consistency of my later work, so unfortunately I would only publish them after refining their tags (at which point it would also make its way into Symix). I could train the current Symix artists individually as Pony LoRAs but I'm not sure if they would be any better than referenced within Symix, and costing a lot more disk space (each artist would probably be around 90MB as a Pony LoRA). It's something I can consider.

>I could train the current Symix artists individually as Pony LoRAs but I'm not sure if they would be any better than referenced within Symix, and costing a lot more disk space (each artist would probably be around 90MB as a Pony LoRA). It's something I can consider. Nah the current ones are fine and work perfectly within Sysmix. It's just the missing ones that come up all the time and aren't there and people goo "aww...". Stuff like miramira, trinityfate, Jaykuma etc that never even made it into the medium let alone high category. There are people out there that still to this day, that's what they wanna see and only request those artists. And yeah I could pull out the old SD 1.5 and do em but ehhhhh... I don't think I can go back...

>>24570 >Jaykuma, miramiraclerun Both are in the current version of Symix! You may have missed it because I don't advertise it very heavily. They're in the artists wildcard text file in the current version for Pony. Trinityfate was waiting on someone else to help me with tagging but that fell through so I'll have to get to it later. There are others that haven't made it in, I just haven't gotten to them yet.

>>24571 Oh wild, I see Jaykuma listed now on civitai but not miramiraaclerun. How daft that I never just... tried them. I had at one point just copied all the artists into a dynamic wildcard and never checked again, but Jaykuma is staring me right in the face. But anyway, cool! That was two big ones that people were still clamoring for, so great to know I can do them. *edit oh ok I see now, I was still looking at the super old page of Symix not the new ep27 one, got it. I'll update all my wildcards with the new stuff. Everytime I think I'm on top of things...

>>24572 miramiraclerun is in the artists wildcard file in the "training data" zip, I just checked to confirm that. You might want to check that file to see if any other artist got missed in the various version upgrades. I'm trying to provide documentation as a zip per version to minimize confusion.

>>24542 >>24553 Looks like there's no PonyXL LoRA for Mipha, but there is one for Urbosa. Anywho, have some more Zelda I generated a while ago. I posted some of these in the Bottom Heavy thread.

More SDXL experimentation. I've been skipping symix because it lowers gen quality too much and I don't find I need it to get size

>>24587 these are so good, thanks so much, I hope you make more too

>>24566 Just wanted to quickly say thank you for all the hard work you've put into this! I was skeptical of Pony (didn't help I had to jump to ComfyUI in order to make it work, which is a whole other ballgame), but now that I've been playing with it, it + Symix has been an absolute revelation. So seriously, thank you so much.

>>24764 I have largely been ignoring AI art until now and i'm really impressed that it's gotten to the point where you can now request specific Homestuck trolls and it looks like them

>>24838 Yeah its called a Character Lora. You run through the AI on a set of images specific to a character and it will create a dataset that you can pop into your prompter of choice and helps give really consistent results. IF you want to see Loras in action, there's a bunch of examples in the image resquest thread.

Random selection of Mongirls I've genned

And some goblins, Spoilered for preg.

I might have a thing for skinny women to gain weight and transform into hucows.

I think there is a lot left to optimize in my generations.

JK Heavy Cruiser discovers an often-overlooked downside to compensated dating: all those lavish, quadruple-digit price tag meals have to go *somewhere*... and that somewhere is her, naturally.

Of course, now that she's putting the 'heavy' in 'heavy cruiser', she's attracting a whole new set of clientele. Which only exacerbates the problem.

>>24565 Huh? Am i missing something? One link contains a singular image and the other is dead? Am i regarded or just blind?

found these on DA and whatever model this guy is using must be pretty good

Does anyone have the art from this game called Fataclysm? The only way to access the art is the play through the game, but I was curious if anyone has gotten all the art from it already. https://jupitarian.itch.io/fataclysm

>>25396 The assets are encoded in the html so I had to manually extract and save them. This took me 2 hours. https://files.catbox.moe/kskrb9.zip The barley mobile and immobile stages are on the supporter patreon build so can't extract them with the public build.

>>25418 Thank you very much! If you want the link here has the patreon build, which includes barely mobile and immobile pics. aHR0cHM6Ly9tZWdhLm56L2ZpbGUvRTNGMzJUWVIjN0hqSDg0WjJNZTYzVFlyM0Fua09uT1JKVVo5ak9ZY2dxYnBENUdKdkFtOA==

>>25422 You do you decode these links?

>>25422 How do you decode these links?

>>25422 Sweet, thanks. This should be everything now and a bunch of new assets I found. https://files.catbox.moe/g818tn.zip

>>25427 Thank you very much for your work! >>25426 The jumbled mess is in base64. Copy and paste it in a base64 translator to get the link.

>>25009 Your goblins are great, I'd love to see more, and bigger.

>>25484 Thanks! Here are some more, I'm trying these out with an immobile lora I found on civit. Seems to give the flabbier bellies that I prefer.

>>25527 you bastard, i will never fully recover from these images

>>25513 Damn, those are very nice. I'm sure there's a lot of cool stuff you can do with your program. Always like to see ones in the vein of your second pic, where their size is inconvenient in some way. Like using multiple chairs, needing machines to move around, getting stuck in doors, clothing too small. Fat's awesome, but it's always improved with a visual reference, you know?

Link here: https://www.deviantart.com/pffedup

>>25534 Yeah, that particular lora is pretty nice for the really massive sizes, but the AI still struggles with generating those extra points unless you want to invest time inpainting those types of details in.

>>25642 You have the loli board, freak, go on post on there.

I really like v5 of SyMix. All the issues I had with it seem to be gone. Plays well with multiple concept and style lora

Back with more of coffee shop girl getting fat. >>25638 This one is fantastic, nice work!

And the follow-up, where she gets even more fat & bloated. And also got a chance to visit the beach!

>>25756 >>25757 Nice to see your coffe shop cutie again! You’ve really perfected the ball belly look, has the perfect amount of weight and bulge to it. And that beach pic ballooning out of the shorts and bikini top might just be one of your best yet man, top marks

fat bump of fat girls in various styles

>>25925 nice kronii

it's stupid how good ponyxl is

>>25994 Need to refine these but seem good enough to post too

Trying out a ufotable lora for thinner 'before' pics. Might try it with a multi-pass workflow at some point since ufotable doesn't have a whole lot of great fat training.

Her level isn't the only thing she maxed out.

>>25995 Those are great.

>>24007 would love to see more of this gal

>>26088 I'll see if I can recreate her. My workflow has changed drastically since then and also there was an update to that checkpoint.

>>26199 no worries if you can't

Hyperfusion v9 for PonyXL is now available. Really good at beeg sizes. https://civitai.com/models/655844

>>26933 It's really good, but I would recommend using ADetailer for faces. Hyperfusion tends to mess up faces pretty easily.

>>26933 I did. Generated at 1024x1024 but only upscaled by 1.25x because I wanted to quickly upload something. Doesn't help when the faces are *tiny*.

https://www.deviantart.com/thepffedup/art/1090978333

https://www.deviantart.com/thepffedup

>>27144 Just buy an ad.

some random gens. i need to start refining them and figure out an actual workflow. if anyone has styles they'd recommend in PonyXL i'd love to hear them. also looking for a good way to gen videos from images.

>>27152 Why'd girl number 5 put a fruit gusher on her nipple?

Been running some favorite images from other creators through img2img in ponyrealism.

yahna got big

meguca

>>27452 Reminds me of some of my local police chiefs

>>27783 Not a Dragon Maid guy but I like that Avacado was able to get these to look authentic

>>27801 That's me! I do my best to find a style that matches the character when generating my images.

Some random gens.

If I hadn't prompted this myself, I would not have been able to tell that it was AI generated.

any good?

I remember a few years ago people would put their drawings in IMG2IMG and then laugh about how badly they turned out. Obviously this isn't perfect but...damn. I really think we're gonna end up at a point where AI assisted art either becomes the industry standard. Or it could end up like steroids in sports: a situation where everyone's using it, but nobody owns up to it.

>>27885 BRRROOOO you win the AI award this year hahah this a classic!

idk what thread would be appropriate to ask but does anyone have oardanaree's artworks?

>>27967 https://bbw-chan.link/bbwdraw/

Can someone make a lora of Sharlia from One Piece

A fat, bloated gyaru. My PC recently died, and while I've gotten the WebUI working again, it's not generating *identical* results anymore; probably either some bit of configuration file nonsense, or maybe just hardware differences. Quality's about the same, at least.

Have a random fat goth, because we need more of those

>>28103 How did you get this art style?

anyone got mangobuoi art?

Trying out some big fat asses. Not the best, but it's at least something I've not done before.

What's a good fat hentai model for sd 1.5? I'm still using abyssorangemix and it's not really working out.

>>28132 If you want a 1.5 model that is trained on fat art I recommend Symodel and BiggerGirls. You don't need to use a LoRA alongside it, just look at each model's tags. BiggerGirls has variants for IRL / semirealism as well as anime stuff, whereas Symodel is better tagged so you'll get what you want with less abominations but almost exclusively for 2d stuff. BiggerGirls: https://civitai.com/models/2664?modelVersionId=85646 Symodel: https://civitai.com/models/175874/symodel-bigger-girls-variant Hyperfusion is also an amazing model that does more than just fat stuff. True to its name you'll get great hyper sizes. Hyperfusion: https://civitai.com/models/186849/hyperfusion-finetune-14m-images For generic models I never delved too deep into hentai models. I mostly used AingDiffusion, AnyLora (and it's Anime Mix variant called AAM) as well as one called Manmaru Mix for a slightly less traditional anime look. Taking a look through my old stuff I guess tif you want something close to what you've posted you could use CetusMix or HelloFlatAnime. CetusMix: https://civitai.com/models/6755/cetus-mix HelloFlatAnime: https://civitai.com/models/102893/helloflatanime

>>28131 For some early attempts these are still amazing, especially that third one! You’re really nailing all the gyaru fats.

>>28112 I have no idea actually but I think because when I add fishnet stockings it's trying to add like a panty and stocking-esque artstyle to it. I got it to do it again but only after I added the fishnets.

Does anyone know how to prompt the AI so it can make clothes with multiple colors like this

Hey gamers. Haven't touched SD in ~4-6 months now, so i'm a bit outdated and out of touch. Currently using forge webui with autissmix_pony. Basically what i wanna know is what the current lora + model meta is. I'd prefer sdxl models but if theres been a new really good SD1.5 model i'm all ears thanks in advance lads

>>28330 Your preferred Pony model variant + the latest Symix lora off civitai.

It's been a fat (hah) minute since I did anything with this girl. Still trying to get a workflow that I like for PonyXL since all my old favorites are SD1.5 based. Most of it is probably too vore-centric for this thread, but I've gotten a few working. (bonus cloning accident in one of 'em lol) I've still gotta figure out how to use the SyMix thing that replaced all the individual artist LoRAs.

oops, this was supposed to be #6 in that one

Hope the body doesn’t look too weird

>>28414 While it is weird, it’s somehow very sexy. I especially love the first one.

Here the normal size

>>28431 Interesting, not a lot of artists do that sort of bloated upper-belly look, despite in nominally being more realistic. Looks unique!

I know she's practically a stick figure by this board's standards, but I've managed, after much wrangling of inpaint sketch (two-tone hair was a terrible decision), a pic of my FFXIV character after spending a little too long with the well fed buff. Though she's apparently not too unhappy with the results.

Miss Hattie AI WG dump

Here for some of genshin character (Improve some prompt from elysia one)

Got a few more of her, with a much bigger gut this time.

I've been slowly working my way to getting better pictures. Started putting the good ones on pixiv and deviantart for fun and preservation. Here's a few.

>>28675 care to drop that deviantart link tho?

>>28676 Sure thing. https://www.deviantart.com/grandpashiddenstash

Midjourney is pretty mid at making fatties but this wasn't too bad.

Some schoolgirls

How about some Dutch girls?

I heard the Dutch had all girl armies that ate people throughout their history. They are so interesting. It fascinates me how Dutch are such strange things. >>28648 These are so cute.

More of coffee shop chick's SDXL variant getting fat.

>>29536 Damn, she’s cute every time you make her Also some of the best ‘leaning forward’ poses I’ve seen, every time I try a pose that isn’t just standing upright it comes out looking wrong. Great stuff!

Could anyone make some Nahida chubby images? Or vore images? Thank you.

Fat -ADVENT- that im posting here because they dont fit into any of the 30+ ai threads on the chan.

>>29576 That last one! That GUT! Magnificent. And the face is adorable too. If this character was one of my blorbos I'd be fuckin losing it but it's still a really great image.

Bit less rounded belly in these, but I've been *really* pleased with how the saggy belly LoRA comes out. >>29551 The secret is to generate 100 images and only keep the ones that look good lol. That, and Pony seems to be pretty good at poses with relatively little extra prompting, as long as it's not too rare of a pose. >>29699 oh man, highschool me had so many fantasies of doing that first image to Tali...