First day taking a crack at this. The possibilities are terrifying. Makes me want to make a "Fattening Models" Hearthstone-style cardgame, with all the assets AI generated. Message me if you like cardgames, or want to help me figure out some prompts and shit together: biskitson#8777

really liking the fat artist style loras. the bws and sizemologist ones are working well for me. looking forward to more! hoping for a miramiraclerun one! >>1684 you can use the png info tab in the a1111 web ui

Is this the place to ask if there is a Loras of cookies-cat?

>>2541 Ah, the first baby steps. Have fun homie! I will note, your gens have the tell-tale sad sepia/faded tones of not having a correctly setup VAE color correction file. VAE setup and some sample VAEs are listed in >>3 Also, about the card game, had a similar idea. Maybe look into Dulst (https://dulst.com/). Could certainly be something fun to collab on. Everybody makes 3 characters and two spells, 25 people later you've got a card pool goin'.

I made some cuties today

>>2542 Could I get a catbox of those Mikus please? >>2545 Also request a catbox of the 2nd, 4th and 6th images here please. Great stuff.

>>2542 Thank you, sorry, i'm newbie in this

>>2541 can you upload these to catbox? >>2542 can you upload the maya one to catbox?

What do you think?

Does anyone know any fat face/chubby face Loras or prompts?

>>2562 Not sure how blobby you're trying to go, but here's a few examples of just adding stuff like "double chin chubby face puffy cheeks neck rolls slob" somewhere near the front of the prompt. I figure you could probably adjust the level of fat faceitude with gimp's warp tool and a little img2img, if you're looking for extremely fat faces.

Does any have a Lora/model for the artist Jeetdoh?

>>2560 Another example

>>2565 What was the prompt you used of the first 3?

>>2569 masterpiece, best quality, highres, 1girl, (portrait of obese massive attractive teenage woman with green eyes double chin fat face puffy cheekys), (red hair), (befuddled expression), (fat breasts), (freckled cheeks) Negative prompt: (worst quality, low quality:1.3), (depth of field, blurry:1.2), (greyscale, monochrome:1.1), 3D face, nose, cropped, lowres, text, jpeg artifacts, signature, watermark, username, blurry, artist name, trademark, watermark, title, multiple view, Reference sheet, cowprint, ((bikini)) Steps: 25, Sampler: DPM++ 2S a Karras, CFG scale: 8, Seed: 1552865105, Size: 512x512, Model hash: e26eec3f09, Model: BG70+AOH30, Variation seed: 1016548735, Variation seed strength: 0.6 masterpiece, best quality, highres, 1girl, (portrait of obese dimpled attractive teenage woman with green eyes double chin fat face puffy cheeks big jaw), (red hair), (languid expression), (spherical breasts), (freckled cheeks) Negative prompt: (worst quality, low quality:1.3), (depth of field, blurry:1.2), (greyscale, monochrome:1.1), 3D face, nose, cropped, lowres, text, jpeg artifacts, signature, watermark, username, blurry, artist name, trademark, watermark, title, multiple view, Reference sheet, cowprint, ((bikini)) Steps: 25, Sampler: DPM++ 2S a Karras, CFG scale: 8, Seed: 3562594315, Size: 512x512, Model hash: e26eec3f09, Model: BG70+AOH30, Variation seed: 2559895178, Variation seed strength: 0.6 masterpiece, best quality, highres, 1girl, (portrait of obese heavy attractive teenage woman with green eyes double chin fat face puffy cheeks big jaw), (red hair), (spiteful expression), (vast breasts), (freckled cheeks) Negative prompt: (worst quality, low quality:1.3), (depth of field, blurry:1.2), (greyscale, monochrome:1.1), 3D face, nose, cropped, lowres, text, jpeg artifacts, signature, watermark, username, blurry, artist name, trademark, watermark, title, multiple view, Reference sheet, cowprint, ((bikini)) Steps: 25, Sampler: DPM++ 2S a Karras, CFG scale: 8, Seed: 2450119369, Size: 512x512, Model hash: e26eec3f09, Model: BG70+AOH30, Variation seed: 1043984680, Variation seed strength: 0.6

>>2567 Yo is that Alisae? (Or however you say her name xd)

>>2572 Yes, I tried to see how good can be merged loras on characters and fat artists’s loras and models

What do you think? model: deliberate_v2 CFG scale: 9 seed: 1150854095 positive: full shot of a cute woman (heavy bottom:1.4) (large thighs:1.6) (fat belly:1.2) (big breasts:1.3) (chubby:1.2) (obese:1.2) (busty:1.2) (thick:1.2) wearing a red small dress and has short brown hair negative: (((no clothes))), (((nude))), (((naked tits))), ((((naked boobs))), (((nipple))),(((breast exposed))),(((topless))), (((naked))), (((young))), ugly, duplicate, morbid, mutilated, extra fingers, mutilated hands, poorly drawn,(lazy eye), (heterochromia), lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad feet, extra limbs, (multiple navels), (two navels), (creases), (folds), thin, slim, athletic, muscular, fit, blemished stomach, rash, skin irritation, poorly drawn face, blurry

Anyone want to fatten these? It’s an interesting aesthetic for them to not just be drawn larger, but for them to exist in a world where they already were, something easiest to show in merchandise.

POV - you caught Kongou snacking at work (no, I have no idea why the window is doing that, couldn't get the solid half to function like an actual window)

>>2581 One more, no clue whats going on with the background

>>2577 Realistic models continue to get better and better, yet still, are uncanny enough to be mostly not aesthetically pleasing. >>2579 No idea who these characters are but I took a stab at it. Just a bit of upscaling and pre-img2img warp tool, no inpainting to fix hands and such. Bit of openpose for the hat character.

>>2581 >>2582 Could you upload those to catbox please?

>>2583 Serval and Kaban from Kemono Friends. I was honestly hoping more this >>409 than resembling direct reference redraws, at least there was an attempt.

>>2585 The second photo. I didn’t particularly want to include it in my post because of the character, so I’ll link it instead since it’s the only thing like it.

>>2586 Still not sure...what you mean. Do you mean... you want these physical toys to be depicted fatter, but, still depicted as physical, plastic toys in the real world?

>>2587 Presented as if actual objects to give off some sort of walking into an alternate universe feeling, yes, since with episode screenshots it’s significantly easier to just edit them by hand since the characters are 3D Models behind 2D backgrounds there. In comparison, it’s much harder to retain the genuine plastic feel. It’s uncanny as hell.

>>2588 No idea then. You'll need to ask that anon how he got that super plastic action figure look on that one picture you mentioned: I can't get anything quite like it.

>>2588 Well, actually, I got a few halfway decent gens from protogen model (anime version). Could you? Sure. Should you? ehhhh (professional photo of a detailed hyper obese woman yellow hair fox ears action figure scenes), (sleeveless white shirt), (leopard print skirt), (yellow bow-tie), (closed mouth smile), (yellow eyes), made of plastic, doll, toy, (long fox ears), (white gloves), wide hips, fat arms, lot of details, detailed, intricate, masterpiece, 4k, volumetric light, hdr no upscaling or inpainting so they could have looked a little better, but honestly the plastic thing is kinda creepy so I just left em at 512x512

>>2548 https://files.catbox.moe/127o7y.png https://files.catbox.moe/6e9gpd.png >>2552 https://files.catbox.moe/ejc2yh.png >>2579 I love Serval and Kaban, so here. I've been using a pvc lora from the touhou ai discord. there's a bunch of good stuff posted there. otherwise, I think there are some pvc styles on civitai

so far my only non-abomination

>>2593 Big thank you

>>2581 >>2582 Fantastic, can I get a Mutsu or two it you don't mind?

anyone know if there's a bellyfuck / navel insertion LORA yet?

>>2598 sure, any particular scenarios in mind? obviously nothing too complex

>>2602 Nothing really cones to mind, maybe looking shy about being so plump

>>2604 Here you go, damn shy fatties are cute

I sat in a bumplocked thread for two days like a moron, wondering why it got so slow. whoops. Glad to see we got the board, though >>2539 Very nice stuff on the guide thread. All that's missing is a list of extensions

>>2609 So nice, thank you!

>>2611 could i get a catbox for that sex pirate?

>>2616 Sure https://files.catbox.moe/ddxben.png >>2617 >>2608 > desaturated > grainy > bad eyes > purple splotches Pretty good chance you don't have a vae enabled, get one if you don't have one from >>3 , put it in your models/VAE folder, and select it from settings

>>2620 thanks for the link. hate to be greedy but could i get that first giant marine ass too please?

>>2621 https://files.catbox.moe/ggjwe7.png > hate to be greedy You're fine

>>2623 thanks again, fantastic work as usual. i wish we had the /vtai/ extension here.

Surely someone doing the vtubes do Choco Sensei plz

how are you guys getting such good hands? inpainting and luck?

>>2629 Model, inpainting, luck. Some models are inherently more detailed then others but a lot of it is pure luck and inpainting

>>2629 I used bad hands, bad anatomy, missing fingers, extra fingers, twisted fingers, double hands in negative. If I have time, I make edit in ps or csp

noticing that most pics use the "1girl" prompt, but what do you change if you were trying to make 2 characters have sex or something?

>>2635 >noticing that most pics use the "1girl" prompt, but what do you change if you were trying to make 2 characters have sex or something? Use img2img and make sketch, it can be very simple

>>2635 You could try 'latentcouple' which is an extension you can download to more easily generate two subjects in an image, be that people or objects

>>2634 Could you catbox that first image pls?

>>2642 https://files.catbox.moe/eptukw.png

>>2642 if you want you can change blue eyes on brown eyes to make Ameliance Leveilleur

REPOSTING FROM PREVIOUS THREAD: Here it is. The merged bbw models. Pic related is a comparison between all of the combinations I made using a prompt that some might use with this model. VAE is the Anything v3.0 one. I think going with a more saturated VAE is probably a better choice. Kuki needs more work, the model I merged with needs less aggressive weighting on certain blocks of the unet so that bbw-step isn't diminished too much. https://gofile.io/d/AzRqHS (All of the models here in the above link need to be pruned. Most of the models in the above link are here, except for the full sized Cocoa only model. ) Some general notes regarding the model: The model is based on Anything V4.5, which means that this should be compatible with most LoRAs and embeddings. If you want to get backgrounds that are even more detailed, using the token "detailed background" will activate dpep3 much more aggressively. As for negative prompts, go for whatever you would use for an Anything v4.5 or AOM2 based model. Some notes regarding the Cocoa and Tea Merges: These two have some oddities introduced from Cocoa which causes the masterpiece keyword to create some weird results, although not consistently, so experiment with it in your prompt and with it not in your prompt. I have had good results with and without it. Additionally, make sure your clip skip is set to 2, otherwise you might encounter some oddities. Recipe for the merged model First, merge dpep3-chillout into bbw_step_30000k (when I say this, you should make bbw_step model A, and dpep model B) (find it here: https://huggingface.co/closertodeath/ctdmixes/blob/main/dpep3-chillout.safetensors ) with the following Merge Block Weights (you cannot use the default checkpoint merger, use supermerger, or another extension that can do MBW) 1,0.9,0.7,0.5,0.3,0.1,0.4,0.4,1,1,1,1,0,0,0,0,0,0,0,0.1,0.3,0.5,0.7,0.9,1 (You can start from the half version in the above link if you want, the fp16 and normal version did not have any hash differences) Then merge the previous model you just made, I would call it bbw-dpep3-chillout, with Cocoa at the following block weights: 0.15,0.15,0.15,0.15,0.15,0.31,0.3,0.3,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15,0.15 (NOTE: FOR ANY MERGE INVOLVING SPECIFICALLY THIS MERGED MODEL, ONLY DO MERGES WITH THE FULL SIZED MODEL AS THE HALF SIZED ONE HAS A DIFFERENT HASH FROM THE FULL SIZED ONE) Then merge the previous model you just made, I would call it bbw-dp3c-Cocoa, with Tea at the following block weights: 0.35,0,0,0,0,0,0.1,0.1,0.1,0,0,0,0,0,0,0,0,0,0,0,0,0.35,0.35,0.35,0.35 (You can find Cocoa and Tea here: https://huggingface.co/andite/desserts/tree/main Cocoa is most likely a mix of YohanMix and Blossom-extract, but I have very little information on Tea ) Anyways, let me know what you think, I feel like the merge managed to really fix a lot of the issues the base model had. My favorite is either cocoa or tea, I haven't done enough prompting with either to really come to a conclusion. I'm probably also going to experiment with merging HLL at some point as well. ADDENDUM: I have made additional merges involving different recipes for chuubas. If you are interested in that sort of thing, you can find it here: https://mega.nz/folder/v80QhCCZ#TaTb_8vErCq0efZYdFh35Q

>>2645 Thank you

Good prompts for inpainting weird shaped belly?

>>2584 Apologies about that 3 day wait, missed your message till I scrolled back to find something https://files.catbox.moe/mbrd9k.png https://files.catbox.moe/h6jd7y.png

https://gofile.io/d/Ef3Pg6 - made a lora for TrintyFate's style was pretty rushed (just used deepbooru tagging) but can give decent results, I found it works best at about .35-.5 weight. Hope you guys enjoy

>>2653 Oh wow, this is spot-on their style. I recognized it immediately! I certainly miss when they used to draw human stuff (since nowadays they only draw furries). Are you able to generate any pics from the other side in panties? Or larger sizes? Still impressed by how well those four turned out though!

>>2656 it can generate pics from the back pretty well, this is about the biggest girl i could get out of it. i imagine incresing the weight would let you get bigger but usually results in messed up faces. i trained it and am using a mix of bbw 30k steps and counterfeit at about 70/30 iirc so use that for best results maybe training it on only bbw 30k might also give bigger girls, might give it a go in a couple days

>>2660 Trinity was really good at drawing ass, and it seems the AI knows that as well haha. These turned out well, minus of course a bit of oddness around the outline of the characters on a few pics, which is expected of an AI. And no worries, I know larger sizes are pretty difficult, especially considering (to my knowledge) Trinity drew more thick stuff than they did immobile or blob.

>>2635 https://github.com/opparco/stable-diffusion-webui-two-shot if you want two specific characters you will need latent couple to generate them. But if you don't care if one or both of the characters are generic, you don't need it and can prompt like normal and use "1girl, 1boy, sex, NSFW, penis, [position/act of your choice eg doggystyle]" >>2649 you'd probably be fine to just use your normal prompt with high-ish denoising (0.65-0.8?) >>2653 Holy shit nice. It's scary accurate. would you mind sharing your training config?

Even on Bing image creator can get some damn good images

>>2593 May I ask which model you used for the last one?

>>2673 What prompt you used for that one

>>2673 This one is amazing, nice work

>>2675 NTA but get in the habit of asking for catbox uploads instead of prompts. Negatives can often be just as important as positives and you will want to have all the information you can get when trying to recreate things.

>>2664 apologies but i don't really have a config, i'm not very computer literature so just followed some reddit tutorial on how to make loras using dreambooth and used a premade colab sheet. i've provided the settings that i think are relevant but again sorry if they are useless

>>2675 +slightly shy, slightly embarrassed, slightly nervous, blush, wet skin,masterpiece, best quality, highres, (button gap:1.1),teacher, nude,(huge belly on desk, fat:1.2), (tits exposed:1),brown eyes, in a classroom, shy, fat ass, fat thighs, chubby arms, brunette, -less than 5 fingers per hand,more than 5 fingers per hand, deformed hands, deformed fingers, too many shadows, worst quality, low quality, extra limbs, extra body parts, ugly, uneven eye, deformed, deformed face,heterochromia,short hair, (green border), (bottom green border), multiple navels, dark skin, too many shadows Also used this model and one of its example images to base this picture off (picrel) https://civitai.com/models/11141/biggergirls-anythingv4-buns-and-thighs

>>2690 forgot one

>>2690 can you upload the second image to catbox and also >>2691 this one too?

>>2692 https://files.catbox.moe/pzkes7.png https://files.catbox.moe/9axv2q.png hope i did it right lol

>>2694 thanks!

>>2689 That model is the same bbw30k that has been floating around, but uploaded by a rando. Just saying so people don't go and download something they already have (it also might not even be the most recent version) >>2685 The lora is great and it looks especially good when you prompt more detailed lighting/rendering & backgrounds. trinity's style is too simple otherwise imo. I also appreciate the settings, thanks

>>2703 Can you please box images 3, 4 and 5?

>>2704 Sure https://files.catbox.moe/g02yds.png https://files.catbox.moe/dpdb97.png https://files.catbox.moe/u20z2o.png

>>2706 Thanks

>>2653 I hope that you make a guide to training loras, because this is pretty amazing work.

>>2653 Would you be able to share a screen shot of your settings? I can't figure out how to get a good picture to save my life, all mine come out like weird droopy aliens.

>>2545 can you share the model you've used in the six cuties? Thanks

Has anyone tried making any tomboys, I’d love to see them

>>2718 thanks, i had just followed a guide (look up PlumishPlum lora) for it myself so i don't really have any secret knowlegde to give you. i took around 90 pics and resized them to 512x512 and auto tagged them using train tab on the webui >>2735 here are the settings i use mostly, after generating a good image i will use highres fix to regenerate at twice the res, my model is a 70/30 mix of bbw 30k steps and counterfeit with a anything v4.0 vae

>>2664 I'm pretty new at this and am having trouble getting anywhere close to my desired results. Could you tell me what models and prompt were used for that fifth image?

Is this the place to ask if there is a Loras of cookies-cat?

>>2543 >>2758 there is none. yet. feel free to make one

>>2755 If you want those results you should make upscale after generations

AI art has its own board now? That's nice. Anyway, here are a bunch of artist LoRAs: https://mega.nz/folder/sShj1LpS#Zh5ZxGO3uwK4QEaZ2X6z1w >>2758 I have the data for it, I might train one at some point.

>>2764 Oh, looking back to previous threads, it seems that someone already posted my mega link from /vtai/. I stopped posting here for a while because of the incessant hostility from anti-AI posters.

>>2767 welcome back. hopefully that will stop now since we have our own place. also requesting catbox links for 2, 3, 4 and 6 please.

>>2768 here you go friend https://files.catbox.moe/v8pnte.png https://files.catbox.moe/iolqmp.png https://files.catbox.moe/wnu4xt.png https://files.catbox.moe/th7aox.png

>>2755 My model, short instructions, & some examples are here: https://mega.nz/folder/WFk13apa#WvCEO8IoqgSaeoKQILeIqg You could also (and probably should) use the model from >>2645 , which is the same model but a cool anon merged it to fix some of the issues with faces & backgrounds That image you mentioned's prompt was "(masterpiece, best quality, highres, detailed:1.1), 1girl, mori calliope, red eyes, pink hair, tiara, panties, bra fat, wide hips, thick thighs, fat hips, fat thighs, sweaty, in a fancy room, dramatic lighting" with negative prompt "(worst quality, low quality:1.4), blur, blurry, depth of field, motion lines, (bad background:1.1), (flat color, sketch, 3d, cartoon:1.1), border, skirt, hoodie, dress shirt,button-up shirt, outside, (steam:0.9), dress, (fat legs:1.1), (skin folds:1.2)" But you should get in the habit of asking for catboxes instead, because you can put the catboxed image in SD's 'PNG Info' tab to get all of its info (which saves us from these horrible, messy prompt dumps). Try it with this catbox: https://files.catbox.moe/yrmjh0.png >>2764 >>2767 it's nice to see you (and your posts), and thanks again for your work on the loras

>>2775 thanks anon; you also make some great posts, thanks for being around.

>>2775 Ah, I was wondering why people were asking for the catbox when it just seemed like regular image hosting. Thanks!

>>2686 How?

>>2774 thank you for the links. >>2776 could you catbox images 2 and 3 please? fantastic stuff

>>2782 voila https://files.catbox.moe/eyzgbq.png https://files.catbox.moe/6jtfv1.png

Finally Presi will be replaced. I'm so happy

>>2785 thanks again

>>2788 I'm so happy, that this idiot will be replaced. YES!!!!!!!

>>2542 Added miramiraclerun LoRA https://mega.nz/folder/sShj1LpS#Zh5ZxGO3uwK4QEaZ2X6z1w

>>2799 God I love the fats of allisei x3 (sorry if I misspelled her name xD) she doesn't get as much love as she should lol

>>2804 Agree, she and her beauty mommy should have more love

>>2798 What was the prompt you use?

Had some spare time and decided to mess around with the local version of SD shared here on the forum. Jesus christ fatty models + loras are incredibly powerful. Only played around with Midjourney and Dall-E before but manage to get consistently OK results. After adding some textual inversion models, whatever the hell those are, stuff really took off. Question: I've had an issue where some pictures will turn out blurry, almost as if taken by an out of focus camera. Looks like an over the top depth of field effect. I've tried using negative prompting which seems to work somewhat, but is there a way to consistently avoid it? Also, can blurry pictures be fixed with upscaling? And in this case, which model is best suited for "upscaling" to remove the blur? >>2643 Stole this Anon's prompt for a base btw. Works really well. Thanks for sharing.

>>2816 For this art did you use new version of bigger girls?

>>2817 No, I dont think so. I am using: https://civitai.com/models/3748?modelVersionId=4163 Which AFAIK is at least partially based on it. I haven't been able to get the model posted in the guide thread to provide satisfying results.

my creations :)

>>2813 That's so great

Takamoom lora for you. https://mega.nz/file/kh12RIYZ#jy_w1uRRDQn8yRnWYvLRBavvRyqGKfWL1n2fy4BFD5w Works best around a weight of 0.4 to 0.6, but will depend on your model.

>>2852 Can yo make presi lora? Thank you

>>2852 Based.

>>2703 is this a jaykuma or a trinityfate model? I can recognize the style Wonder if my commission from him was fed into the AI, dont care honestly

What do you think about alternative reality where Himeno didn't die and was fed by demons until immobility

A comparison of the most common up scaling methods, (Hi-res fix) vs (img2img SD upscale), and the various upscalers.

what settings are you guys using for your LoRAs? i tried anything v4, v5, and trying to do stuff with the bambooale lora at like 1 strength just results in some wacky tacky stuff

>>2867 Yeah, it's a trinityfate lora from >>2653 , that anon did a great job on it. And since bbw30k and the lora were both trained on trinity, there's a pretty decent chance your commission is in there twice actually, lmao >>2846 Glad you like it, you got some cute dragons yourself. Try using highres fix as well, 720x512 is too small >>2896 Depends on lora & model, I know the trinityfate lora shouldn't really go above 0.4 strength so you could try somewhere around there > i tried anything v4, v5 If playing wiht the weights doesn't work, try using a fat model, most normal models have no concept of fats and do them pretty bad even with a lora

>>2889 good work on this, is this going in the guide? >>2825 kek

>>2897 That smug kronni is great, could you box 3 and 5 pls? Also any updates on when your new model will be ready if you don't mind me asking?

>>2897 specifically it was a jaykuma pic

It's hot, no ?

can you upload the second one to catbox?

>>2685 >That cow girl is great! Any chance of a catbox link please?

New model merges! I have created a huggingface for easy use with colab and other notebooks. Find it here: https://huggingface.co/mariaWitch/biggerModelMerges HLL4 is weird. I'm not sure how I feel about it, it's better and worse at the same time. >bbw120kg 6.0+1.0 You Can (not) Lose Weight This is a fun model to use. This was originally supposed to be a punishment model for the /vtai/ colab, but colab anon found something better instead. However, I think it is still all right. Recipe is as follows: 1. bbw-30000k-step+(biggergirlsv3-NAI)*0.4 2. Take results from one, and follow previous MBW recipe up to cocoa, I was lazy and used half models for this entire merge. 3. result from 2 + (7th_HLL438_v2_AC_WS25 - NAI)* (0.35 or 0.4 I genuinely don't remember, which ever matches 08603ca4e4 is the correct one though) >bbw150kg 6.1+1.1 You Can (not) Lose Weight HD Take the previous model and add tea following the original mbw recipe for the tea model.

>>2918 Also of note, I am going to upload bbw5.0+1.0 which will be a non holo mix, step 2 of the model I described here.

>>2918 Nice work as always. I saw the bbw_hll4b2 botans in vtai and they looked pretty good, I want to give it a try later today >>2905 https://files.catbox.moe/tzk2x5.png https://files.catbox.moe/x18q98.png > when will the model will be ready Maybe a week or two? I know it's taking a while but I usually don't feel like doing more work when I come home from work, sorry. I'm sure you can sympathize > update The new train was overall pretty bad and I need to make a lot of changes. Sure, it got backgrounds better, and if you spoke gently (and used a lot of inpaint) you could coax some interesting images out of it, but the vast majority of the outputs were unusable. If I had to guess, I would say it's because of my inconsistent tags, and I need to go back and reorganize them, which I've been meaning to do anyway what i've learned and what i want to do, briefly: The other anons were absolutely right about the text encoder- training at 15% steps made the model way more flexible, reduced how 'overbaked' it looked despite using the same steps as 30k, and on custom tags gave better results with less training data. The retrain will also have a trained text encoder Since I've last trained, dreambooth can now pull captions from .txt files instead of just the filename. Now, I don't need to have training tags in the filename, I can have them in .txt files which will really help. They'll be easier to edit and I can be more descriptive without worrying about the (really small) filename character limit, which should help quality. I can also now use a normal dataset tag editor which will speed things up That last point will be especially helpful because I have to go back and reorganize tags to fix the outputs. While I'm there, I also want to add a few artist tags, style tags, and size tags for easier control. More poses, better. This should be obvious but I originally left out lots of pictures because I didn't think the ai would 'get' it, which completely defeats the point of training, which is supposed to teach the ai something it doesn't know. There's some more poses now and they've helped very much (check >>2813) and I want to include more Bucketing. Dreambooth can now do bucketing in which it breaks large images up into smaller images, so we're no longer limited to 512x512 pics and can use higher quality imgs for higher quality outputs. I haven't used this feature, nor do I know how to, and it probably wouldn't even be useful since whatever bucketed images I add won't make a difference when there's 1100 other images that have already been cropped and resized to 512x512. But, it's nice to know, and if I get really bored one day I might redownload everything for bucketing Loras are incredible. I avoided them because I thought my dataset would be too big, but then I saw that loras can handle hundreds, even thousands of images perfectly fine, so maybe I should train a lora and training a full model isn't even the smart thing to do? I've been trying to learn how to make them to compare results but I can't get the program to run properly Hope you don't mind the essay, I had a lot to say

>>2927 Sounds like valuable lessons to learn. Hopefully it comes out better on your next attempt!

>>2927 These Botans? I have a great time with this model

>>2927 Yeah ofc man, I feel you, take as long as necessary. Your stuff is always great.

Made some very large bbw giant catgirls, some with slime, some just big. Having a lot of fun with the 150kg bbw model and perspective prompts

Slimy ones

>>2935 The very same. I'm sold on this mix, it's GOOD (and I am absolutely going to feed some of its outputs back into the retrain). Better fats, poses, faces, variety, backgrounds, texture, hands, and I don't even need to use a different model to inpaint. My only gripe is that the extreme ends of bottom heavy are harder to get to and are considerably smaller; (incredibly thick thighs, incredibly wide hips, fat hips, fat thighs, fat legs:1.3) would normally be so much it ruins images but now it only gives the 30k baseline. >>2954 >>2955 > catgirls, slime based. Cute gens, as well >>2937 Liking #2 & #4

what ai are you guys using

>>2975 stable diffusion through automatic1111 webui >>3

How do I check the prompts in the catbox?

>>2980 download the catboxed image, take the downloaded png, then put it in 'png info'. as long as the catbox poster shared the right file it should show you the prompt, gen info, loras, etc used. catbox is the most popular image share but anything that preserves metadata works fine (eg mediafire or mega.nz) if you don't have access to SD or prefer to not use it for whatever reason, you can also use any image metadata viewer

>>2971 Could you catbox 4 and 6 please?

>>2984 👍 https://files.catbox.moe/rcqd62.png https://files.catbox.moe/6zl55c.png

>>2986 Thank you. Out of curiosity, how many of the rest of you are also on /vtai/?

>>2987 I am. I usually get a bit carried away and start making bigger girls that turn into giantess, and then I need a cold shower

>>2988 Go on... And don't afraid to elaborate.

>>2989 Just a lot of stuff like this, and these ones above >>2954 >>2955

>>2988 These were great, especially the okayu korone duo, hope you don't mind the edits >>2987 >>2988 That makes three, so far >>2989 🤨

>>2988 can you box that last botan?

>>2997 https://files.catbox.moe/srie5t.png

Love this trinityfate Lora, made these using it. Going to check out the AllVtubers Lora now. Keep up with the good work. >>2988 Could you box the first image? >>2996 And box the third image of this post, please?

I'm going through the tags now, and the amount that are fucked up is unbelievable. My favorite is NSFW.jpeg, so far. And I also don't recall ever typing "e), e\)" >>3000 https://files.catbox.moe/xgo540.png Nice gens, vaguely reminds me of warioware's ashley >>2993 could i get a catbox for the 1st and 5th image? thanks

>>3036 https://files.catbox.moe/qs5jdm.png https://files.catbox.moe/fxsw4t.png btw whats that tagging thing? >>3000 can you catbox the 1st and 3rd pic?

>>3000 https://files.catbox.moe/k64ery.7z Here's a bunch I made yesterday, mostly pairs, but the okayu might be there too

Back with more beeeeg elira, thank you to the anon who made the new hll merges, they're great

>>3000 Link to the LoRA?

>>3040 can you catbox this please?

Preview :) https://files.catbox.moe/fz00tv.png

>>3040 couple of mios aswell - embarrassed or smug options available

>>3042 sure https://files.catbox.moe/0ka3dc.png

Could someone make some tomboys?

I call this one "mio eats the entire menu" excuse the fucked up hand and double navel, I must have tried 20 inpaints and couldnt get a proper belly grab.

>>3047 sure

It finished. Test Images from all 13 payloads. (Some of these have Dynamic CFG, some of these don't, all of them have been through Hires fix. These prompts were all run on 20 steps DPM++ 2M Karras( with most having 7.5 cfg) when scored) There is 1 non-fat prompt, where I just prompted for a character with no modifiers, and one prompt that should result in an image that isn't fat focused. (Polka and Noel) I may have fucked up the prompts Polka's generation 2 times, the third time I did it properly Suisei: https://files.catbox.moe/izd9b9.png Watame: https://files.catbox.moe/zj0uj7.png Polka 1: https://files.catbox.moe/9yxuo0.png Polka 2: https://files.catbox.moe/i8v713.png Polka 3: https://files.catbox.moe/km66fr.png Noel: https://files.catbox.moe/7ari39.png Mumei: https://files.catbox.moe/jo92in.png Mio: https://files.catbox.moe/cloew1.png Lui 1: https://files.catbox.moe/bjeein.png Lui 2: https://files.catbox.moe/u6xa1y.png Lui 3: https://files.catbox.moe/poink3.png Lui 4: https://files.catbox.moe/llvbl9.png Laplus: https://litter.catbox.moe/gi73mt.png Kronii 1: https://files.catbox.moe/gfqfo1.png (lmao, the scorer was rating how good Kronii's fat butt looked) Kronii 2: https://files.catbox.moe/3ci6lt.png (using DPM++ 2s a Karras at 41 steps) Irys 1: https://files.catbox.moe/gnrtrb.png Irys 2: https://files.catbox.moe/dqnob8.png Irys 3: https://files.catbox.moe/aaqmi1.png Ina: https://files.catbox.moe/80rcfc.png Haato 1: https://files.catbox.moe/e5k7is.png Haato 2: https://files.catbox.moe/1unn0s.png (detailed background added to tags) Ayame: https://files.catbox.moe/pknk3a.png I can't believe Kronii's fat ass infected the seed that I normally use for that Kronii prompt.

>>3050 Forgot images.

>>3050 >>3051 Model Download is here: https://huggingface.co/mariaWitch/ExperimentalBiggerMerges/resolve/main/bbw_hll3.1%20Gods%20Amongst%20Men%20Edition.safetensors Recipe starts with bbw2x5.0+1.0. Use supermerger to merge BBW2x5.0+1.0+ alpha*(hll3-last-pruned - NAI (pruned in ModelToolkit and saved as fp16)) and use the following for the MBW weights (first value is base_alpha/base): 0.7266212993147018, 1.0,0.9033511319969127,0.3208271934949668,0.32212843781144684,0.0,1.0,1.0,0.14938394537570357,0.0,0.37475392081506487,0.30460893915650167,0.742834937815587,0.30450607107638084,0.5959275468063191,0.16121068239626557,1.0,0.8384245305044884,0.7340690362501738,0.29900594232129596,0.16312821599906427,0.0,0.6327754709637872,0.0,1.0,0.24725758816820623 (Note: it may actually be impossible to recreate this model hash with supermerger due to unknown differences between Bayesian Merger and Supermerger causing model merges with the same exact MBW weights and models to not have the same hash.) (Don't forget to tick "Use MBW" in supermerger otherwise it won't work!)

>>3049 >Con you catbox these please?

Has anyone made a Lora for Cookies-Cats yet?

>>3038 >>3041 It's just a merge of 0.7 bbw step 30k pickle import and 0.3 counterfeit V 2.5 with the trinityfate Lora. Here you go! https://files.catbox.moe/jsanly.png https://files.catbox.moe/yd1u29.png

Getting used to prompting with these fat models.

>>3057 Looking good so far, be sure to experiment with different weights to get them in different poses and sizes and stuff

>>3058 Nice, will try to play around with Blender modeling once I'm done getting prompting down. Found this video where you can make your own adventure using Stable Diffusion and language models, this is giving me so many ideas. Not sure if this is more chatbot focus. https://www.youtube.com/watch?v=x5VUidEMDDI

Is there any Loras for huge hips & huge asses?

>>3054 no but maybe if you ask a 5th time >>3041 ctrl+f >>2653

>>3049 sweet lord, catbox for 2&3?

>>3068 >no but maybe if you ask a 5th time Not so fast. What if IIIIII ask for the opposite so that there won't ever be a Lora for Cookies-Cat?

>>3069 The cat's out of the box.

>>3066 There's models for those things here: https://huggingface.co/mariaWitch/biggerModelMerges As an absolute beginner, bbw steps 30k is powerful and user friendly.

>>3054 As anon said in this post >>2754 , this Twitter account has a guide for making your own Loras. The guide seems too be simple, haven't followed it myself yet.

>>3053 >>3069 https://files.catbox.moe/pa0vbj.png https://files.catbox.moe/sgpqox.png

Third /vtai/ ban, let's fucking go. The most annoying part is fulfilling a request and not being able to post it >>3038 it's the dataset tag editor extension for SD. Handy, so I don't need to download a separate program, but it can only work on about 400 images at a time which is pretty annoying >>3052 >>3051 >>3050 Nice job man. I'll give it a spin tonight >>3066 The bbw models are perfectly capable of hips&ass but if you insist on using a lora, one doesn't really exist (yet! ;) ) The closest would be hyperfusion but it's garbage except for big tits https://civitai.com/models/16928/hyperfusion-100k-images There's also atomichips but it doesn't do huge https://civitai.com/models/17376 But really, just stick to >>3073

>>3077 You too huh? I think the janitor on vt hates hags, preg, and bbw specifically

Anyway, here are some big Mios, I'm in love with her at any size but (((bigger))) is always better https://files.catbox.moe/ih9cd6.png https://files.catbox.moe/vqsl6w.png https://files.catbox.moe/25gufv.png https://files.catbox.moe/ca4bbh.png And preemptive catboxes

>>3077 Same fucking here. Banned for an ON-TOPIC post as well mind you.

>>3082 It sucks that so many of us that are actively contributing get bonked for architecture. I post plenty of thin and SFW girls all day long and then get a 24 hour ban for a big Mio

>>3077 >>3080 >>3081 >>3082 >>3083 > Get told fetish content belongs on /d/ > Post fats anyway > Get banned > surprisedpikachuface.jpg And why even post on 4chan at all when this board exists?

>>3085 Have you seen the shit jannies allow? SWF fat vtubers on a vtuber board is incredibly tame by comparison. Posting fat vtubers started because THE VTUBERS THEMSELVES talk about eating or gaining weight. We've been here longer than that low-T fag of a janny anyway. The fact that it gets flagged as "off topic" and not nsfw is a dead giveaway that its just one idiot angry at a fetish he doesn't like. I post the bigger lewder stuff here anyway, and as I already mentioned, I'm posting a lot more than just the occasional thicc vtuber

>>3085 > Get told fetish content belongs on /d/ Not posting fats directly. It is in a catbox, and we still get banned. Even when it is fully SFW.

>>3085 Because I do more than just post fats, and because I want to share the models I merge with more than just the handful of people that know about /bbwai/

You know, I knew a silicon merge was going to score well, I just didn't think it would score this well to start.

Giant goombus rises from the sea! Aparently adding "underwater" to my tags got me some interesting views and a nice heart-shaped butt. Catboxes included for funsises: https://files.catbox.moe/ybp4bd.png https://files.catbox.moe/5cm4v3.png https://files.catbox.moe/pvg5av.png https://files.catbox.moe/xo0xtf.png Goodnight all, I'll be back in vtai after work tomorrow, see you there. The thread always grinds to a halt when we're all vacationed lol.

>>3089 It literally scores better than any of my other models did from the start. For the sake of comparison my newest model, the best composite score my new model got was with it's final model and that was a score of 5.835275851763212

>>3080 >>3077 >>3082 Out of curiosity to my fellow /vtai/ posters, why don't we use the /ag/ thread on /d/? It is for fat vtubers afterall. I know the guy who maintains the lit archive is an anti-AI faggot but we outnumber him and we would be perfectly within the rules there.

>>3077 Could you box that 5th image please?

>>3093 Let's not. It's more for folk who draw by hand. Let's stick to some moral ground

>>3093 Because there would be more people getting mad at AI art. And the amount of {{{bigger}}} chuubas generated is only a small portion of /vtai/s image base.

>>3090 Holy seggs

>>3085 > fetish content belongs on /d/ Oh, the /d/ ai thread? The one that is nothing but futa day in and day out (with the occasional tentacle, lolicon, and horse dick) without any chunky in sight? THAT /d/? Is this bait? > Post fats anyway, Get banned >>3082 This guy's post didn't even have an image and he got banned Not to mention banning fat girls just because they're fat is retarded. Completely covered up, nothing lewd or provocative, 100% sfw content gets you sent to the gulag. Fetishistic? A little, sure. But cunny and feet, among MANY others, go untouched, funnily enough. And it wouldn't even be annoying if the enforcement was consistent, but it's completely arbitrary. Sometimes you're fine to post uncensored porn and other times you get banned for breathing. Or sometimes you're banned for what's more or less the same post you made a day ago or even just a few hours ago without issue. > And why even post on 4chan at all when this board exists? Because most, if not all of the people posting fat girls don't only make fat girls, they have just as much fun making plenty of normal shit and even fulfilling requests, and then don't post to bbwchan because it's, well, not bbw. And then they get banned for posts that don't break any rules and a lot of people aren't even aware of this site >>3093 We'd be within the rules but we should probably leave it to the hand drawn art

>>3095 https://files.catbox.moe/fyg84c.png

>>3085 >And why even post on 4chan at all when this board exists? What >>3099 said. The majority of content I post on /vtai/ isn't even fat. While I used to post fat Mios all the time, I'm more well known for posting cute loli vtubers now instead. While I do enjoy architecture, it's far from the only thing I want to do with AI. I can't just go posting cute little Subarus here or else it's gonna start arguments.

Artificial art is best art.

>>3081 Had some ideas

some i made today

>>3103 Can you please link the trinity fate lora?

>>3105 From here >>2653 https://gofile.io/d/Ef3Pg6

>>3103 Big mio is kino

>>3058 Could you box these two, please?

I am once again humbly requesting Ramlethal Valentine of any size (although these all-around-massive Trinityfate-esque figures are god-tier), or failing that, any tan/dark-skinned fatties with blonde hair you gentlemen may magnanimously generate. Please and thank you.

>>3111 I tried, my models are tuned to vtubers >>3110 Was more or less my first real attempt at 2shot with random characters https://files.catbox.moe/0ldl0u.png https://files.catbox.moe/956f8o.png

Requesting more hologamers, botan and gura huge and wide butt arigatogosaimas

how hard would it be to get a character in another characters outfit Like ill request tamamo from fate in kiara's outfit for example, could any of you do that

Wpuld love some laplus and shion with gigantic ass

>>3117 Would something like this be agreeable? Not a rear shot I know, but wide hips and thighs.

>>3118 Your batch was kino anon. I was there when janny wiped your first link on vtai. If youd able to use the suisei pose to get shion and laplus then i'm die thank you forever. Great stuff

>>3114 That's like, all I do now. Will do more later

>>3121 UUUOOOHH SEEGGS

>>3120 I will get on it.

Also got some more Vtuber / Hologamer pics, kind of messed up tk's hair, hope ya'll don't mind

>>3123 Adding on botan and gura if its not a hassle. Arigato in advance

>>3124 All chubas good aslong as bbw

>>3111 had no idea who ramlethal was but I am now a fan. 'All-around-massive Trinityfate-esque' Ram + a chunky alt >>3112 catbox for 3&4? >>3116 It's hard to get the outfit right even on the original character, let alone get it right enough for cosplay. But if this https://files.catbox.moe/bchpjl.png is accurate enough for you then it works for us, just say the word

>>3127 NTA, but damn thats a tasty looking ramlethal

>>3123 I tried to mix up the locations, but it didn't really work out besides the last one. There is some minor size differences which is good. Also, clearvae is singing my colors, which sucks. It might be time for me to remove monochrome from my negatives. Gura was genned with blessed2 vae as a test.

Random blue wolf/fox girl I default to making into a bbw for practice. Would call it an OC if it weren't AI >>3127 https://files.catbox.moe/n2con1.png https://files.catbox.moe/z2p5cg.png

>>3125 I linked to the wrong post. Your request was fufilled in this post. >>3134

Architected some OL Fauna today, I'm hoping you fellas will like it. Wouldn't post anything bigger than this in vtai for fear of bonks, but I wanted to share anyway

>>3134 >>3136 I am but a mortal in the presence of divinity Thank you forever Would botan and guras tail be difficult with your method? Genuine question.

>>3134 Just an aside for a moment, but the fucking hair on this model is SO FUCKING GOOD. It has the anime aspect closer to the scalp, but as you move further away the hair gains depth, texture, and becomes more strand like. It's probably the best fucking hair I have ever genned, and it was the model merge basically doing all the work. >>3138 I didn't specifically prompt for their tails, so they weren't included. But the fact that they didn't come with them when their name was inputted means that the way the angle is being done the model probably doesn't have a good understanding on the rear shots for Gura and Botan and just treated them as humans. Let me try and regen them with shark tail and lion tail added to both prompts respectively.

>>3139 Nice, la+s tail looks difficult i aint gonna ask you to but you try if you want to www And yeah the hair looked so good i had a hard time thinking it was genned

Gura's tail is basically out of the question. Without significant inpainting it's going to end up looking like tenticles, or have a weird appendages sticking off of it. Botan on the other hand, could work, but using the lion tail token forces a front POV on almost all seeds. I'll continue working on Botan. But I am giving up on Gura.

>>3142 Its a good start Guras tail might be viable in the future so put that on ice. Thanks for your service

Pretty much my first time making these, made with dezgo.com btw

>>3144 Really liking the three pairs of boobs squished together, might have to steal that idea, hehe

>>3143 Unrelated Ayame. But I got something with Botan that I like. It isn't great but it should do.

>>3134 Could you box those please? >>3142 This second gura, >>3146 And this Botan as well?

>>3146 You're getting there with the tail Ayame looks so good

>>3147 That should be all of them. https://files.catbox.moe/7sbbt0.png https://files.catbox.moe/n83wja.png https://files.catbox.moe/xl50kl.png https://files.catbox.moe/coyapn.png https://files.catbox.moe/yij7nt.png https://files.catbox.moe/ld0grb.png

>>3149 Fantastic flabby asses, thank you.

we eating good tonight. the vtai mass exodus was the best thing to happen to this board

Is the anon who posted the Elira stuck in the doorway in here too? Did you ever post the model you were working on with the stuckage and facesitting? I was really looking forward too it.

To everyone here: I have finished the FINAL (until bbw model anon comes out with a second version of the model) version of bbw-hll3.1. It is in the huggingface repo, if you know where to find that. The model is significantly better at everything that bbw_hll3.1b2 could do. I HIGHLY recommend that you try it at least once if you are using bbw_hll3.1b2 or bbw_hll4b2. I will be thinking of some ways to spruce up a model that doesn't have vtubers in it for those that just want to gen only generic anime in the coming days, but bbw_hll3.1 Final should be able to do good generic anime in a pinch. https://huggingface.co/mariaWitch/biggerModelMerges

>>3151 Didn't really leave, I've been hanging out here and there for quite a while

>>3155 Embarrassed mio love Any chance on some big butt aswell?