/abdl/ - Adult Baby - Diaper Lover

For Lovers of Diapers and Ageplay!

Index Catalog Archive Bottom Refresh
Options
Subject
Message

Max message length: 12000

files

Max file size: 32.00 MB

Total max file size: 50.00 MB

Max files: 5

Supported file types: GIF, JPG, PNG, WebM, OGG, and more

E-mail
Password

(used to delete files and posts)

Misc

Remember to follow the Rules

The backup domains are located at 8chan.se and 8chan.cc. TOR access can be found here, or you can access the TOR portal from the clearnet at Redchannit 3.0.

US Election Thread

8chan.moe is a hobby project with no affiliation whatsoever to the administration of any other "8chan" site, past or present.

(134.81 KB 768x1024 6.jpg)

(104.84 KB 768x1024 7.jpg)

(161.45 KB 768x1024 8.jpg)

(131.13 KB 768x1024 9.jpg)

(168.10 KB 768x1024 10.jpg)

I trained a LoRA for SDXL using ~400 diapergal pics as a dataset, all with girls wearing Tena diapers and the results are fucking amazing: Infinitely much better than whatever you get from SD1.5 or any other LoRA, and with much more consistent quality. This is the fucking future bros.
(149.80 KB 768x1024 11.jpg)

(109.38 KB 768x1024 12.jpg)

(160.32 KB 768x1024 13.jpg)

(145.89 KB 768x1024 14.jpg)

(150.42 KB 768x1024 15.jpg)

Here's the LoRA: https://mega.nz/file/mYQjSTqK#mIOmUjvT7ohFKJTEuNozFk664_QtNpQ03f-atk0f0nQ I made it with the Kohya LoRA Trainer Colab for SDXL with 10 epochs and mostly the default parameters (I set "cache_text_encoder_outputs" to True). That's the first version and I think you could get even better results if you experiment with the training parameters, unfortunately my PC is kinda crap so it takes me a lot of time to try stuff. If you want to try training it yourself here's the dataset I made: https://mega.nz/file/mAgjBBTa#FQG9BABZnWBCuxEljG0hhDxgKKl_ctcG2xfl0ihSCH8
What are the specs you use to train sdxl? I am able to train 1.5 on my 8gb 3070 but am curious if my card can train sdxl.
>>29822 I used Google Colab which gives you an Nvidia T4 with 16 GB of VRAM. From what I've read it is possible to train SDXL on 8 GB but I think it needs some optimizations first.
>>29802 Did you use a classic optimizer or one of the new auto adaptative like Prodigy? It theoretically save a lot of time by not having to tweak the learning rate but each time I tried it for SDXL results are so slow it get ridiculous (like 50 hours to train a 3000 step Lora,WTF) Just curious if I’m doing something wrong or if this is expected.
>>29840 I used this colab: https://colab.research.google.com/github/Linaqruf/kohya-trainer/blob/main/kohya-LoRA-trainer-XL.ipynb All settings set to default except for "cache_text_encoder_outputs" which I set to True since it said it was recommended for SDXL. It took like 2 hours to train on Google Colab.
>>29841 Thanks for that, I have a megamax lora cooking as we speak. Your lora is great, but it really prefers that hands on hips pose, which makes sense since most diapergal pics look the same.
(1.17 MB 1024x1024 00050-339043474.png)

(1.49 MB 1024x1024 00045-4141109267.png)

(1.16 MB 1024x1024 00049-3390898047.png)

(1.05 MB 1024x1024 00034-4272888843.png)

(1.33 MB 1024x1024 00042-1951834945.png)

>>29856 This is the result of ~150 images trained on xl based nintendo 6.0. I'm training another with more epochs to see if it can get the tapes a bit better, but it's still pretty good IMHO.
(1.33 MB 1024x1024 00043-3155644556.png)

(1.17 MB 1024x1024 00033-3925280655.png)

(1.16 MB 1024x1024 00194-2526530595.png)

(1.68 MB 1024x1024 00116-2573080971.png)

(1.35 MB 1024x1024 00148-957782543.png)

Here is a Lora I made using ~240 images. Part 1
(1.05 MB 1024x1024 00167-3807751569.png)

(1.14 MB 1024x1024 00094-622448800.png)

(1.30 MB 1024x1024 00057-615458587.png)

(1.27 MB 1024x1024 00000-3402365621.png)

(1.13 MB 1024x1024 00072-2160140666.png)

>>29875 Part 2
How do you caption this for training? Do you tag the image one by one with an activation word ?
(185.24 KB 1002x1022 00115-264045752 - Copy.jpg)

(177.66 KB 1000x1024 asdasd - Copy.jpg)

(160.60 KB 1024x1024 00004-2291931818 - Copy.jpg)

>>29878 Use BooruDatasetTagManager. It has a bunch of preset tags. Just make sure for diapers you have the same tag in each caption. I usually use "wearing diaper"
>>29880 someone should try belle delphine
>>29911 I've seen delphine naked nothing to see there. I mean, literally nothing to see there
(2.24 MB 1280x720 delphinedips.mp4)

>>29931 Did you make this?? Only other deepfakes I’ve seen are from diaperedextreme and would love to see more.
first of all hopefully we can avoid pedo shit this time, i am pretty close to never visiting here again second of all i got these from merging the realistic omorashi model on civitai with a model called aniverse. loras were: <lora:ABDL:0.15> <lora:DiaperMixLoRA:0.9> <lora:InCribLora:0.2> <lora:lora_trainer:1.0> <lora:weight_slider_v2:0.75> 'lora_trainer' is diapermess v0.4. most of these are on huggingface in a repo called abdl pack, very googleable. clip skip 2-3 and MAKE SURE TO PUT (ch_ld, l_li, sh_ta, under_ge, p_do:1.2) in the negative prompt!!! (replacing underscores with the expected vowels of course) or else things will get fucking disgusting and finally, you can open any lora in notepad to see what tags it was trained on. diapermess v0.4 for example uses "diapermess" but the omorashi checkpoint uses "messed diaper." i used (huge messed diaper:1.5), (gigantic diapermess:1.5) to get most of these.
(2.40 MB 1840x920 grid-0119.png)

(2.30 MB 1840x920 grid-0091.png)

(2.37 MB 1840x920 grid-0101.png)

(2.40 MB 1840x920 grid-0090.png)

(2.44 MB 1840x920 grid-0115.png)

(3.44 MB 1280x720 gadotdips.mp4)

>>29940 Yes I made it in Stable Diffusion. Here is an attempt at Gal Gadot
(16.47 MB 1280x720 Margot Robbie.mp4)

>>29978 Pretty damn good. Always love to see these. I’ll contribute one Of the few I have collected
(1.19 MB 1024x1024 00017-1484283832.png)

(1.06 MB 1024x1024 00363-611714304.png)

(1.08 MB 900x1024 00339-2227873530.png)

(1.19 MB 1024x1024 00202-194628173.png)

(1.13 MB 1024x1024 00178-924433171.png)

>>29955 >>29957 >>29958 >>29959 Damn quite a few good ones in there, thanks.
Any that aren't femoids?
(176.44 KB 978x978 OIG.nnfF.jpg)

(155.48 KB 978x978 OIG.jpg)

(181.15 KB 978x978 IMG-20231003-WA0002.jpg)

How can I use the tenaXL-v0.safetensors file?
>>29931 wow looks cool, any pics?
(1.23 MB 1024x1024 00009-2073108861.png)

(1.16 MB 1024x1024 00022-887804275.png)

(1.14 MB 1024x1024 00013-160998660.png)

(1.11 MB 1024x1024 00017-1432496472.png)

(1.41 MB 1024x1024 00018-1121625358.png)

Someone uploaded a half decent sdxl diaper lora on civit. Faces all kind of look the same, but otherwise it's ok.
(159.09 KB 1024x1024 OIG.S2mu48emd3vOjQT.jpeg)


(145.34 KB 1024x1024 OIG (1) (21).jpeg)

(1.18 MB 1024x1024 00040-2205164835.png)

(1.17 MB 1024x1024 00122-1531292336.png)

(1.13 MB 1024x1024 00049-248777748.png)

(1.21 MB 1024x1024 00052-4051626845.png)

(12.77 MB 1152x1536 AnimateDiff-ManmaruMix_v3.gif)

Anyone expirementing with Animatediff?
(522.67 KB 512x880 00096-2035034922.png)

(553.29 KB 512x880 00237-988304729.png)

(557.12 KB 512x880 00241-988304733.png)

(591.78 KB 528x904 00346-3968257673.png)

(643.58 KB 624x920 00360-1710553617.png)

(1.39 MB 1024x1024 00057-3234298816.png)

(1.35 MB 1024x1024 00056-2371845297.png)

(1.32 MB 1024x1024 00044-2806085309.png)

(1.21 MB 1024x1024 00033-2955943228.png)

(1.26 MB 1024x1024 00064-2262432889.png)

Surprisingly SDXL does a fairly good job at creating large baby stuff.
>>29801 Are there any SDXL models that are based on Japanese jojifuku fashion?
Damn I didn't really think AI generated images were there yet for satisfying diaper porn but a few of the ones in here are really amazing. Just being able to get the diapers to look a little bigger and fit to the body correctly and being able to edit some of these scenarios and facial expressions would send this into the stratosphere. I'm not tech inclined. I just use Microsoft Excel all day for work but I kinda want to get into this. Any guides or tutorials anyone would recommend for starting out?
>>30606 >Japanese jojifuku fashion? i'll have to look, but not that i'm aware of any. Definitely would would be a great lora or model. I feel like base SDXL might be able to replicate
>>30647 Hi Aifog, any update on your pixiv? Are you uploading anywhere else? I love your work!
https://www.pixiv.net/en/users/56400 this account has been insane this past week or so and I need to share the news! Alter sin babylook is back BABY!!
(2.09 MB 1024x1536 113174211_p4.png)

(2.38 MB 1024x1536 113174211_p7.png)

(2.23 MB 1024x1536 113174211_p5.png)

>>31067 here's a throuple of kewl cyberpunk cuties that were dropped today
(540.78 KB 512x768 112948701_p0.png)

(563.21 KB 512x768 112948701_p1.png)

(550.72 KB 512x768 112948701_p5.png)

>>31067 I want to watch this anime
(558.32 KB 512x768 112957018_p22.png)

(543.91 KB 512x768 112957018_p9.png)

(512.60 KB 512x768 112957018_p34.png)

(564.54 KB 512x768 112957018_p0.png)

(446.21 KB 512x768 112957018_p38.png)

Mommy! Sorry, Baby! Sorry, Mommy! Sorry, Baby! Sorry...
(502.20 KB 512x768 112846577_p9.png)

(507.94 KB 512x768 112846577_p2.png)

(527.53 KB 512x768 112846577_p11.png)

(453.70 KB 512x768 112846577_p14.png)

Dis one gives me gender envy
(1.08 MB 720x1200 stilldead.png)

(720.00 KB 768x1024 pomni.png)

(973.39 KB 768x1024 aifog (1).png)

(1.78 MB 1152x1536 aifog (2).png)

(3.49 MB 1536x2048 aifog (3).png)

>>31082 nicely done >>30756 well they finally responded, but didn't indicate to me which posting was the problem. I'm just using this as an hiatus. But i'll keep nagging them. Here's some renditions of pomni and otherthings.
(88.41 KB 1024x1024 OIG.47H1LLOPs.jpg)

(116.70 KB 1024x1024 OIG.abdt64x.d.jpg)

(122.33 KB 1024x1024 OIG.jtbrMX.jpg)

(114.50 KB 1024x1024 OIG.SjRkhmb22YQ9W.jpg)

(108.61 KB 1024x1024 OIG.urD_dXFP0quC_7i.jpg)

Managed to get these out of Bing AI / Dall-E 3. Has anyone else had luck generating from it? It's pretty tough with all of the restrictions.
>>31088 What convoluted prompt did you have to make to not having to write diapers? Theses are not bad actually
>>31085 Would you kindly do more Pomni, even better if it's "show accurate" please?
(117.97 KB 1024x1024 OIG.4dMqxFq_TTfhYsV..jpg)

(151.41 KB 1024x1024 OIG.CQQsG.jpg)

(100.68 KB 1024x1024 OIG.drl43ax.jpg)

(119.16 KB 1024x1024 OIG.sddprnrs452.jpg)

(114.57 KB 1024x1024 OIGnthppydprs.jpg)

>>31089 I got a 24 hour suspension unfortunately, so I can't log in now to see my exact prompt, but I think it was: "nurse at a hospital helping an embarrassed female patient with a diaper" This prompt has worked 90% of the time and produces decent results. Sometimes it would work if you add "standing" to the end of it so that the subject isn't always sitting or laying on a bed. Any other modifications to this prompt have been shot down by the system and resulted in a suspension but I'm going to see what else I can get. Apart from this, I got one image out of using: "photo of a woman shopping for diapers wearing a diaper" / "photo of a woman shopping for diapers with a diaper on" So I think that prompt or a variation of it will also work, but I again, I was suspended so I can't see my exact prompts until later. Hoping someone else can give it a try and share their results here.
(1.20 MB 1024x1024 00266-1314938064.png)

(1.04 MB 1024x1024 00281-2357444657.png)

is there a specific site people are using to share their ai work?
Seku new work on pixiv. I kinda love diaper leotard melding that happens in AI porn
Sin https://www.pixiv.net/users/56400 my favorite amateur hentai giving me all the euphoria
Same link >>31150 Look how happy the Boi is!
(1.71 MB 1024x1536 00284-1196196806.png)

(1.49 MB 1024x1536 00275-720764454.png)

(1.40 MB 1024x1536 00264-3150019904.png)

>>31153 what prompt did you use? 'emma watson as a crack whore'?
(1.03 MB 768x1152 00291-847578613.png)

(1.72 MB 1024x1536 00296-2433773489.png)

(1.73 MB 1024x1536 00298-3279085306.png)

(1.76 MB 1024x1536 00307-406785329.png)

(1.81 MB 1024x1536 00315-4156003374.png)

(134.23 KB 1024x1024 OIG.Tpx2vvGs321.jpg)

(118.99 KB 1024x1024 OIG.Z0DJ8d5Hs.jpg)

(86.66 KB 1024x1024 OIG.CUUJd39.jpg)

(185.17 KB 1024x1024 OIG.1jx224s_4kd.jpg)

(129.82 KB 1024x1024 OIG.e__O3DmPQg_7y.jpg)

More from Bing
>>31159 This is so disturbing in the best way. No one working in medicine should be this happy. Love love love!
(95.12 KB 1024x1024 14.jpg)

(102.38 KB 1024x1024 20.jpg)

(109.54 KB 1024x1024 22.jpg)

(108.44 KB 1024x1024 31.jpg)

(109.99 KB 1024x1024 35.jpg)

(95.47 KB 1024x1024 36.jpg)

(114.38 KB 1024x1024 44.jpg)

(128.96 KB 1024x1024 49.jpg)

(130.22 KB 1024x1024 89.jpg)

(137.56 KB 1024x1024 95.jpg)

(113.93 KB 1024x1024 96.jpg)

(109.71 KB 1024x1024 98.jpg)

(128.39 KB 1024x1024 99.jpg)

(130.03 KB 1024x1024 102.jpg)

(122.31 KB 1024x1024 103.jpg)

That's it for now. All from bing. Asking for editorials seems to screw with its content block.
>>31177 I hope You keep this up friend
(127.74 KB 1024x1024 OIG.7lomsKCDTsgRbIP8zTQY.jpg)

(141.26 KB 1024x1024 OIG.TpzQTtYkYexoUqWpdNh.jpg)

(154.98 KB 1024x1024 OIG.i2niXJCa9NrejRiGyx.jpg)

(171.97 KB 1024x1024 OIG.LFn6ySBMlso.jpg)

(176.88 KB 1024x1024 OIG._Oq.tF_DAk_fY4VWy9UZ.jpg)

>>31177 Works great, thanks.
(153.85 KB 1024x1024 OIG.zpAUGJ_aziySvr.jpg)

(157.94 KB 1024x1024 OIG.oXxj70vHZ3.jpg)

(196.32 KB 1024x1024 OIG.9iouku05OaEuKl1Z.jpg)

(179.45 KB 1024x1024 OIG.SLuVItXPR91.jpg)

(127.32 KB 1024x1024 OIG.uiD4IrjCEQ13tg7hD.jpg)

(159.62 KB 1024x1024 OIG.hQRIPQKg6j1ZLS_FiQ.jpg)

(142.90 KB 1024x1024 OIG.025dtD.jpg)

(115.68 KB 1024x1024 OIG.RgiO8TPD.jpg)

(181.28 KB 1024x1024 OIG.WlvKLej7exUYhGX_ru.jpg)

(119.10 KB 1024x1024 OIG.I6CvAm4OshFK.jpg)

(203.79 KB 1024x1024 240.jpg)

(172.53 KB 1024x1024 241.jpg)

(175.59 KB 1024x1024 244.jpg)

(205.12 KB 1024x1024 245.jpg)

(168.05 KB 1024x1024 246.jpg)

Another tip: add 'An inspiring and motivational photograph of...' to your prompts. I dunno, it seems to like it.
>>31217 It's like I have to reimagine What it even means to doodle in the days of blinding mathematics. mesmerizing and dystopian
Been playing around with bing...Can't seem to bypass the word "diaper" getting restricted, any ideas or advice?
I'm struggling getting it to generate full body pictures, always only get like the upperbody...Any advice on what input to use?
AI Story generator I would recommend. Does NSFW and knows what a diaper is https://perchance.org/ai-story-generator
(118.59 KB 1024x1024 OIG.jpg)

(159.09 KB 1024x1024 OIG.S2mu48emd3vOjQT.jpeg)

(104.52 KB 1024x1024 OIG-1.jpg)

(125.52 KB 1024x1024 OIG.lD.jpeg)

>>31252 Nappy seems to work for me in certain scenarios
Which input do you use to get younger adults? I either get children or really old adult
If it's okay to ask, does anybody know of any caches of photos of Apple from ABDreams ideally not all from the same set, the more varied the better
Still struggling with the nappy / diaper aspect, but getting some interesting scenarios out of it
(195.41 KB 1024x1024 OIG.bjEYTchfPOu6zHLK8pK.jpg)

(127.25 KB 1024x1024 OIG.pEfVXwiI04.jpg)

(119.42 KB 1024x1024 OIG.CMz_uvvY6BEVYdH.jpg)

(140.47 KB 1024x1024 OIG.zYGYG7Ge4.jpg)

(181.58 KB 1024x1024 OIG.G7v4LcvGVEpKxCr.jpg)

>>31353 How many images / what model did you train on?
>>31361 Could you do more like pic 3?
>>31362 Seconded
The messy faces are the best. Well done.
1. a guy named tommy goes to college. he has a problem with wetting his pants. he gets called up in front of the class while the waistband of his diaper is sticking out. he is humiliated. his daddy started making him wear diapers every single day after his wet accident in public. he wets his diaper so hard out of anxiety that his diapers start to leak through his pants. the entire class stands up, points, and laughs at him.
(715.94 KB 2880x2880 20231114_060536.jpg)

(892.60 KB 2880x2880 20231114_060455.jpg)

>>31352 >nappy Thanks. I'm using "very gibbous nappy soaked with yellow water" and works half of the times
Anyone tried the new chatgpt text to image feature. Does it also have a nsfw filter?
(1.20 MB 1024x1024 00213-3544368899.png)

had to be done
(392.41 KB 512x768 Amazing!.png)

Pretty sure this is my best one yet
(499.12 KB 2880x2880 20231114_182431.jpg)

(510.68 KB 2880x2880 20231114_182259.jpg)

(525.80 KB 2880x2880 20231114_182323.jpg)

(1.71 MB 1024x1536 00001-2222841518.png)

(1.95 MB 1024x1536 00012-4023078094.png)

(1.81 MB 1024x1536 00051-3180390765.png)

(2.03 MB 1024x1536 00035-618427537.png)

(2.48 MB 1024x1536 00021-966627505.png)

(1.16 MB 1024x1024 00744-3998764665.png)

(1.14 MB 1024x1024 00747-2717867224.png)

(1.29 MB 1024x1024 00752-1811035450.png)

(1.76 MB 1024x1536 00753-3433961802.png)

(2.28 MB 1024x1536 00778-4124864065.png)

(1.84 MB 1024x1536 00766-1241879600.png)

(1.94 MB 1024x1536 00767-2680441543.png)

(2.49 MB 1024x1536 00775-1651093798.png)

>>31402 If you are taking request. ab/DL (heavy on the diaper lover.) street fashion is content I didn't know I needed. Great stuff
(154.42 KB 1024x1024 OIG.GmAPCITx1dUITC.jpeg.jpg)

(142.67 KB 1024x1024 OIG.jpeg-240.jpg)

(181.05 KB 1024x1024 OIG.jpeg-244.jpg)

(145.97 KB 1024x1024 OIG.jpeg-242.jpg)

(158.53 KB 1024x1024 OIG.jpeg-237.jpg)

>fursuit artstyle
(709.14 KB 2880x2880 20231115_213405.jpg)

(737.88 KB 2880x2880 20231115_213545.jpg)

(743.52 KB 2880x2880 20231115_213502.jpg)

(784.73 KB 2880x2880 20231115_213428.jpg)

(87.01 KB 1024x1024 OIG (1).jpg)

(95.44 KB 1024x1024 OIG.jpg)

(123.03 KB 1024x1024 OIG.8q0bI.jpg)

>>31358 i put either an specific age, early 20s or college student
Mind sharing the specific input used for those picture, just to get the idea of what your putting in?
>>31468 Mind sharing the specific input used for those picture, just to get the idea of what your putting in?
>>31474 Sure, for these my prompt was along the lines of "photograph, male early twenties, shirtless, in a playmate in the middle of a nursery, wearing nappy with prints" from there I tinker with hair style, expression, props and such
(1.32 MB 1024x1024 00327-1845270235.png)

(1.49 MB 1024x1024 00370-3011265433.png)

playing with regional prompting so I can do more than just 1girl in [place] and not have everybody in diapers >>31426 I'm trying out new things at the moment, and surprisingly there's very few ABDL stuff out there so I'm making stuff with what I got I do like public exposure too so you'll see that
(681.56 KB 2880x2880 20231116_171740.jpg)

(689.39 KB 2880x2880 20231116_171716.jpg)

(651.89 KB 2880x2880 20231116_185341.jpg)

(669.91 KB 2880x2880 20231116_171811.jpg)

(690.81 KB 2880x2880 20231116_171846.jpg)

Does anybody like these?
I'm slowly getting the hang of the right prompts to get some okay results. Really like to play around with scenarios many of these could be used for captions, I think.
(105.02 KB 1024x1024 OIG (1).jpg)

(105.65 KB 1024x1024 OIG.jpg)

(112.70 KB 1024x1024 OIG.4G.jpg)

Consistently getting four pictures with these for some reason, did the filter get easier for anyone else?
>>31510 Its hit and miss for me. But those you're getting there are great. I think it also easier with male models due to the ability to have bare chest so it gets less restricted and censored
>>31485 I'm interested to know what model you used
(842.02 KB 2880x2880 20231117_202840.jpg)

(751.21 KB 2880x2880 20231117_182139.jpg)

(790.28 KB 2880x2880 20231117_193356.jpg)

(773.81 KB 2880x2880 20231117_180939.jpg)

(810.92 KB 2880x2880 20231117_181339.jpg)

>>31513 [SAFE] the character Rouge the Bat from Sonic the Hedgehog saga, white fur, pink chestplate, black wings, long white gloves and boots with a pink fold, dancing in a Comic Con, editorial, she is wearing a large gibbous cloth nappy that is soggy with yellow water, fursuit artstyle [SAFE] You can also add "multiple similar costumes of" at the start for extra effect
Boy I sure do love seeing this fucking dog over and over and fucking over again even on the tamest fucking prompts.
(605.63 KB 1392x712 fuckyou.PNG)

unfucking believable
>>31514 Which generator bing AI?
(755.80 KB 2880x2880 20231118_052821.jpg)

(738.54 KB 2880x2880 20231118_052844.jpg)

(715.50 KB 2880x2880 20231118_054705.jpg)

(707.10 KB 2880x2880 20231118_054722.jpg)

(712.46 KB 2880x2880 20231118_052908.jpg)

>>31519 Yeah, bing.con/create Also, "inflated gibbous nappy" and "squatting" instead of dancing are good prompts
(851.31 KB 2880x2880 20231118_070959.jpg)

>>31522 "login" No thanks
(124.63 KB 1024x1024 OIG.qrLHNkE.jpg)

(134.07 KB 1024x1024 OIG.rKJEtwPgm.Jm_J62.jpg)

(115.18 KB 1024x1024 OIG.5y_f4Rv3sNja.jpg)

(145.20 KB 1024x1024 OIG.7X0x7U3ZVlB.jpg)

(103.21 KB 1024x1024 OIG.ZkVw3U1.jpg)

My tastes are DL heavy with just women in diapers in various scenarios, but I'm hoping people having trouble can modify these prompts to get images and share their own prompts and generations here. Posted in order: editorial photograph of a woman leaning a bar at a tavern bar, holding a drink, turning and looking at camera, smirk, tattoos, black bangs, cigarette, black tank top, she is wearing a large gibbous nappy that is soggy with yellow water editorial photograph of a stewardess woman wearing a gibbous diaper with yellow stains on an airplane walking down the aisle and serving drinks on an airline, stewardess uniform magazine article featuring editorial photograph of a pouty 20 year old woman wearing a diaper and lying with a pacifier, there are large robot arms holding baby powder, the article says "AI Takeover?" magazine article featuring an editorial photograph of an embarrassed college student woman wearing a backpack and diaper with pants around ankles in a hallway with lockers, pointing college students laughing, the article is titled "Dealing with Incontinence" editorial photograph of a nervous 20 year old woman wearing saggy diaper standing on stage and telling stand up comedy jokes behind a microphone stand at a comedy club, tomatoes being thrown on stage
Anyone done any with an Elle Fanning-like appereance? Reckon she would look cute
(126.95 KB 1024x1024 OIG.JqkVoxENO96Ov.jpg)

(224.97 KB 1024x1024 OIG.g73TqlWm.jpg)

(161.40 KB 1024x1024 OIG.Hm9NpRDy0vTJ.jpg)

(229.03 KB 1024x1024 OIG.rEdQyH0U.jpg)

(149.24 KB 1024x1024 OIG.uRJH.jpg)

>>31536 editorial photograph of a bored 20 year old woman wearing saggy diaper holding a controller and playing video games in a messy apartment magazine article featuring editorial photograph of an diapered 20 year old woman who is homeless holding a sign that says "Will Work for Changes" with diapers on the ground, the article is titled "Homeless In Need" magazine article featuring editorial photograph of a pouty blonde 20 year old woman wearing a diaper and lying with a pacifier, there are large robot arms holding baby powder, the article says "Innovations in Automation" inspiring and motivational photograph of a 20 year old woman doing cartwheels in a public park wearing saggy diaper and having fun, central park with skyscrapers in the background An inspiring and motivational photograph of two concerned college women at the movie theater watching a movie and eating popcorn, they are wearing large gibbous diapers yellow stain
>>31545 i like that first one, it's like "Pampers Grande, the diaper you only have to change weekly"
I didn't expect to get this
WTF
>>31547 We need more ai generated saggy bean bag diapers imo.
>>31549 you cracked it >>31536 I can't believe I never thought of rocket bye baby robot hands
would it be bad optics to guardrail I dunno where it falls on the PC spectrum of weird sexual stuff
(53.42 KB 540x540 48.jpeg)

(156.18 KB 1024x1024 099.jpeg)

(210.63 KB 1024x1024 OIG (2).jpeg)

(237.93 KB 1024x1024 OIG (3).jpeg)

(119.26 KB 1024x1024 OIG (4).jpeg)

>>31550 I decided to add huge/massive to their prompts before diaper and it worked surprisingly well
Any tips on getting tab diapers instead of pull ups?
>>31559 What model and prompts did you use this looks pretty good
(1.17 MB 1024x1024 00214-2272572254.png)

(1.13 MB 1024x1024 00257-3713408143.png)

(1.22 MB 1024x1024 00226-2614403917.png)

(1.31 MB 1024x1024 00222-779926064.png)

(1.32 MB 1024x1024 00219-3611021838.png)

(1.25 MB 1024x1024 00266-930883685.png)

(1.08 MB 1024x1024 00298-4249483296.png)

(1.21 MB 1024x1024 00291-2418822326.png)

(1.24 MB 1024x1024 00297-4249483296.png)

(1.32 MB 1024x1024 00271-1936108625.png)

(1.06 MB 1024x1024 00329-490960122.png)

(1.36 MB 1024x1024 00307-2518060860.png)

(1.38 MB 1024x1024 00311-2518060860.png)

(1.21 MB 1024x1024 00340-1646807630.png)

(1.25 MB 1024x1024 00334-388114475.png)

(1.15 MB 1024x1024 00343-1626008391.png)

(1.28 MB 1024x1024 00365-1800432831.png)

(1.21 MB 1024x1024 00368-1659699635.png)

(1.22 MB 1024x1024 00376-69679499.png)

(1.32 MB 1024x1024 00369-842806297.png)

>>31570 Just Bing "male, college student, shirtless, wearing nappy with prints, inside a playpen, vacant eyes, nibbling on a big teething toy"
(144.91 KB 1024x1024 OIG.jpg)

this is alot of fun
(114.08 KB 1024x1024 OIG.jpg)

(109.63 KB 1024x1024 1.jpg)

(56.80 KB 810x810 2.jpg)

(110.31 KB 1024x1024 OIG.VL0rx8SsgiPFKIQ.jpg)

some more from me
(203.46 KB 1024x1024 OIG.mz0GaJ8D2j9z.jpg)

(206.07 KB 1024x1024 OIG.WCxnFTFrMw.jpg)

(205.73 KB 1024x1024 OIG.eJNpLXWXugY.jpg)

(226.31 KB 1024x1024 OIG.Cn9uERAP4Qeo.jpg)

(132.28 KB 1024x1024 OIG.f_Zm01TFlwX6J6L4.jpg)

>>31545 article featuring editorial photograph of a mother helping 20 year old embarrassed daughter change her nappy, college student daughter lying on her back on a busy public bathroom changing table, pack of diapers on the counter, women laughing pointing at them in background, the article says "Dealing With Incontinence" same for #2 magazine article featuring editorial photograph of a mother helping 20 year old daughter change her nappy, college student daughter lying on her back on a busy public bathroom changing table, pack of diapers on the counter, the article says "Dealing With Incontinence" magazine article featuring editorial photograph of a mother helping 20 year old daughter change her nappy with yellow stains, college student daughter lying on her back on a public bathroom changing table, pack of diapers on the counter, the article says "Dealing With Incontinence" magazine article featuring editorial photograph of a young woman with arms crossed wearing a yellow stained diaper laying a couch in the living room, an angry mother is helping her, the article says "Dealing With Incontinence"
>>31199 Holy shit I love Dipchicks. Let's bring this back.
>>31628 they messed with the filter for sure, can't get any of the prompts I used for these to pass anymore
(172.10 KB 1024x1024 21.jpg)

(175.21 KB 1024x1024 23.jpg)

(171.60 KB 1024x1024 OIG..jpg)

(139.55 KB 810x810 OIG.bPB.jpg)

(120.80 KB 810x810 OIG.jpg)

(208.18 KB 1024x1024 2.jpg)

(181.34 KB 1024x1024 OIG.jpg)

(201.43 KB 1024x1024 3.jpg)

(117.11 KB 1024x1024 OIG.jpg)

(175.18 KB 1024x1024 OIG.Z_U6.jpg)

(146.13 KB 1024x1024 4.jpg)

(154.08 KB 1024x1024 5.jpg)

(153.40 KB 1024x1024 OIG.jpg)

(119.20 KB 1024x1024 OIG.saVecZpn2KZ..Yie.jpg)

(136.58 KB 1024x1024 OIG.ASJz9e7ZRM9wts.Rtd.jpg)

Does anyone know a good model for soiling / underwear pooping?
(115.81 KB 1024x1024 OIG.n7bEeV8Kpux.4.jpg)

(99.73 KB 1024x1024 OIG.eOqUzv65wwl.jpg)

(191.20 KB 1024x1024 OIG.OyO.jpg)

(108.20 KB 1024x1024 OIG.sc44FcR9ruIgh25F.jpg)

(132.06 KB 1024x1024 OIG.YklYAgbv8X6.jpg)

>>31604 editorial photograph of a well dressed man taking his 20 year old woman date out to the prom, balloons decorations and uplighting, the woman is embarrassed and wearing a gibbous nappy editorial photograph of an embarrassed 20 year old woman wearing saggy diaper soggy with yellow water, standing on stage and telling stand up comedy jokes behind a microphone stand at a comedy club, tomatoes being thrown on stage by an angry audience magazine article featuring an editorial photograph of a smirking college student woman with a diaper with pants around ankles in a public bathroom stall looking sitting with index finger over mouth shhh gesture, the article is titled "They'll Never Know" editorial photograph of a well dressed man taking his 20 year old woman date out to the prom, balloons decorations and uplighting, the woman is pouty and wearing a gibbous nappy editorial photograph of a concerned 20 year old woman wearing saggy diaper who has short hair and tattoos cleaning a messy apartment with a duster
(181.63 KB 1024x1024 OIG.vjbiJaRB.jpg)

(164.18 KB 1024x1024 OIG.Dgjs6QIJUtC.jpg)

(220.54 KB 1024x1024 OIG.yXPa9I.5jn.jpg)

(166.09 KB 1024x1024 OIG.iZP2W_Py.jpg)

(197.28 KB 1024x1024 OIG.ncth6d9cB_q.jpg)

>>31676 magazine article featuring editorial photograph of a fast food worker woman spilling a tray of fast food inside of a busy restaurant, the waitress is shocked and wearing a gibbous nappy, customers sitting nearby laughing magazine article featuring editorial photograph of an embarrassed college aged fast food worker woman walking and spilling a tray of fast food inside of a busy restaurant, the waitress is wearing a gibbous nappy, customers sitting nearby laughing, the article is titled "Accidents Happen" (same prompt for 3rd image) magazine article featuring editorial photograph of a college aged fast food worker woman holding a tray of fast food inside of a busy restaurant, the waitress is embarrassed and wearing a gibbous nappy soggy with yellow water, customers sitting nearby laughing, the article is titled "Accidents Happen" magazine article featuring editorial photograph of an embarrassed college aged fast food worker woman walking and spilling a tray of fast food inside of a busy restaurant, the waitress is wearing a gibbous nappy, customers sitting nearby laughing, the article is titled "Accidents Happen"
(185.85 KB 1024x1024 OIG.OgZPt.jfu5E.jpg)

(199.37 KB 1024x1024 OIG.eFT8yIw1.jpg)

(166.41 KB 1024x1024 OIG.Q0P7.jpg)

(159.87 KB 1024x1024 OIG.RgohL3mDkiJ.jpg)

(183.92 KB 1024x1024 OIG.zhdahSEt.jpg)

>>31677 editorial photograph of an embarrassed college aged woman with a standing with hands in the air standing on the path, she is being searched by a tsa agent at a busy airport, the woman is wearing a gibbous nappy, there is a large pack of adult diapers on the xray conveyor belt, people standing nearby laughing editorial photograph of an embarrassed college aged woman with a standing with hands in the air standing on the ground, she is being hand searched by a tsa agent at a busy airport, the woman is wearing a gibbous nappy, there is a large pack of adult diapers on the xray conveyor belt, people standing nearby laughing editorial photograph of an embarrassed college aged woman with a standing with hands in the air standing on the ground, she is being touch hand searched by a tsa agent at a busy airport, the woman is wearing a gibbous nappy with yellow stain on it, there is a large pack of adult diapers on the xray conveyor belt, people standing nearby laughing (same prompt for 4th image) editorial photograph of an embarrassed college aged woman with a standing with hands in the air standing on the path, she is being hand searched by a tsa agent at a busy airport, the woman is wearing a gibbous nappy, there is a large pack of adult diapers on the xray conveyor belt, people standing nearby laughing
(190.87 KB 1024x1024 OIG.dFH5gOUs.jpg)

(135.23 KB 1024x1024 OIG.Ppsod98d7Edhj.jpg)

(188.73 KB 1024x1024 OIG.MLYlh.jpg)

(132.43 KB 1024x1024 OIG.TIkTcfBuvd.jpg)

(153.48 KB 1024x1024 OIG.G7dMaOYUEsC.jpg)

>>31678 editorial photograph of an embarrassed college aged woman with a backpack standing next to a baggage claim with arms crossed at a busy airport, the woman is wearing a gibbous nappy, there is a large pack of adult diapers on the baggage claim, people standing nearby laughing, side view (same prompt for 2nd image) editorial photograph of an embarrassed college aged woman with a backpack standing next to a baggage claim at a busy airport, the woman is wearing a gibbous nappy, she is reaching for a big pack of adult diapers on the baggage claim, people standing nearby laughing (same as 1st prompt) editorial photograph of a college aged woman with a backpack standing next to a baggage claim at a busy airport, the woman is embarrassed and wearing a gibbous nappy, she is reaching for a big pack of adult diapers on the baggage claim, people standing nearby laughing
>>31679 love the third one, it looks properly wet, more scenarios would be good, it's funny like she hopes they don't see the diaper luggage, not like they couldn't see the one that she's wearing
i wish there was a way to use 80s diaper pics to get the more vintage plastic look
(127.09 KB 1024x1024 OIG.jpg)

(160.69 KB 1024x1024 OIG.jpg)

(157.24 KB 1024x1024 OIG.jpg)

(129.29 KB 1024x1024 OIG.jpg)

(166.71 KB 1024x1024 OIG.jpg)

(136.66 KB 1024x1024 OIG.jpg)

(149.06 KB 1024x1024 OIG.jpg)

(172.66 KB 1024x1024 OIG.jpg)

(164.90 KB 1024x1024 OIG.jpg)

(121.18 KB 1024x1024 OIG.jpg)

(179.27 KB 1024x1024 OIG.jpg)

i feel like this AI tool has made my early wank fanasies come true.
(128.07 KB 1024x1024 OIG.jpg)

(146.37 KB 1024x1024 OIG.jpg)

(139.64 KB 1024x1024 OIG.jpg)

(162.61 KB 1024x1024 OIG.jpg)

(131.42 KB 1024x1024 OIG.jpg)

(107.53 KB 1024x1024 OIG.jpg)

(101.54 KB 1024x1024 OIG.jpg)

(113.03 KB 1024x1024 OIG.jpg)

(149.56 KB 1024x1024 OIG.jpg)

(2.06 MB 1168x1560 00173-3411544029.png)

AI, learn how to pour a beer. For fucks sake, look at that foam!
(42.44 KB 540x540 OIG.jpg)

(40.91 KB 540x540 OIG.jpg)

(131.23 KB 1024x1024 OIG.plAz2bMR.hVGus3Wqd.jpg)

(128.17 KB 1024x1024 OIG.jpg)

(122.72 KB 1024x1024 OIG.jpg)

(108.05 KB 1024x1024 OIG.jpg)

(111.10 KB 1024x1024 OIG.jpg)

(59.22 KB 540x540 OIG.jpg)

(163.32 KB 1024x1024 OIG.jpg)

(156.54 KB 1024x1024 OIG.9vx3GSOD57C.jpg)

(151.18 KB 1024x1024 OIG.jpg)

(138.39 KB 1024x1024 OIG.jpg)

(163.32 KB 1024x1024 OIG.jpg)

(156.54 KB 1024x1024 OIG.9vx3GSOD57C.jpg)

(151.18 KB 1024x1024 OIG.jpg)

(175.42 KB 1024x1024 OIG.jpg)

(136.15 KB 1024x1024 OIG.jpg)

(138.84 KB 1024x1024 OIG.jpg)

(147.11 KB 1024x1024 OIG.ccwhVzpn9uM8Csh5wu.jpg)

(155.72 KB 1024x1024 OIG.FLR_Yc6qL.jpg)

(135.50 KB 1024x1024 OIG.jpg)

(173.19 KB 1024x1024 OIG.jpg)

(140.91 KB 1024x1024 OIG.jpg)

(152.58 KB 1024x1024 OIG.jpg)

(151.03 KB 1024x1024 OIG.jpg)

endless trash
(152.98 KB 1024x1024 OIG.jpg)

(143.86 KB 1024x1024 OIG.jpg)

(157.83 KB 1024x1024 OIG.Y1kpwwn3_uK.jpg)

(129.92 KB 1024x1024 OIG.jpg)

(166.35 KB 1024x1024 OIG.ge.jpg)

(117.53 KB 1024x1024 OIG.jpg)

(170.41 KB 1024x1024 OIG.jpg)

(126.04 KB 1024x1024 OIG.jpg)

(132.96 KB 1024x1024 OIG.jpg)

(160.35 KB 1024x1024 OIG.jpg)

(116.71 KB 1024x1024 OIG.jpg)

(178.26 KB 1024x1024 OIG.OJmTVVwWm.jpg)

(143.04 KB 1024x1024 OIG.gKHUZKcKpv.jpg)

(133.50 KB 1024x1024 OIG.VdVCczsaj8dSe1Ahb8.jpg)

(144.02 KB 1024x1024 OIG.jpg)

(137.97 KB 1024x1024 OIG.jpg)

(138.55 KB 1024x1024 OIG.jpg)

(151.79 KB 1024x1024 OIG.jpg)

(187.40 KB 1024x1024 OIG.jpg)

(141.47 KB 1024x1024 OIG.jpg)

(189.64 KB 1024x1024 OIG.Y78Y3L.jpg)

(199.42 KB 1024x1024 OIG.xjZNEpJ8MQ3z4t.jpeg.jpg)

(169.58 KB 1024x1024 OIG.jpeg-1566.jpg)

(176.66 KB 1024x1024 OIG.jpeg-14.jpg)

(155.29 KB 1024x1024 OIG.jpeg-17.jpg)

(164.69 KB 1024x1024 OIG.jpeg-1788.jpg)

POV: Your relative's new incontinent girlfriend shows up at Thanksgiving to confidently argue politics. They break up after this. What do?
(1.88 MB 1432x800 Final_00009.mp4)

SDV is pretty fun, once you get it figured out, though it's a dice roll whether or not the animation will focus on the character or just do a zoom out or pan. Randomly got clouds. This works on Img2video, I also have a workflow one that i can use lora's and any model with.
(2.82 MB 1432x800 Final_00014.mp4)

>>31811 This one is kinda grainy, here's a better example.
(148.56 KB 1024x1024 OIG.jpg)

(182.94 KB 1024x1024 OIG.jpg)

(169.88 KB 1024x1024 OIG.jpg)

(195.14 KB 1024x1024 OIG.jpg)

(157.41 KB 1024x1024 OIG.jpg)

(119.12 KB 1024x1024 OIG.jpg)

(165.11 KB 1024x1024 OIG.w.OBZM0F_Jt8LACeg.jpg)

(143.27 KB 1024x1024 OIG.jpg)

(3.47 MB 1432x800 Final_00020.mp4)

(274.94 KB 922x978 O.jpg)

Hey, hey people. I broke it.
(146.38 KB 1024x1024 OIG (3).jpeg)

I am utilizing ai to make cute little anime guys in humiliating situations. >No way daddy! You can't tell my friends about Baby Bottle Bunny Bib Highchair time. It's bad enough you still make to use this dumb chair just cause I'm short. If they know they won't let me stay in the adventurers guild. >No I am not fussy cause I'm wet again! I hardly ever even wet when awake!
This is without upscaling btw, the one on the left, i have no clue but got a good laugh.
(155.42 KB 1024x1024 OIG.jpg)

(175.80 KB 1024x1024 OIG.iW1p3MEGC53oAbLCy.jpg)

(132.92 KB 1024x1024 OIG.jpg)

(106.65 KB 1024x1024 OIG.TO.jpg)

(131.50 KB 1024x1024 OIG.jpg)

(118.10 KB 1024x1024 OIG.jpg)

(129.04 KB 1024x1024 OIG.jpg)

(129.36 KB 1024x1024 OIG.jpg)

(112.38 KB 1024x1024 OIG.jpg)

(105.44 KB 1024x1024 OIG.jpg)

(119.20 KB 1024x1024 OIG.jpg)

(118.83 KB 1024x1024 OIG.cytIwlYu.jpg)

(108.54 KB 1024x1024 OIG.jpg)

(127.12 KB 1024x1024 OIG.ZlhZBfH38.vhbZTO.jpg)

(164.53 KB 1024x1024 OIG.jpg)

(163.30 KB 1024x1024 OIG.jpg)

(158.11 KB 1024x1024 OIG.o.jpg)

(142.05 KB 1024x1024 OIG.jpg)

(149.70 KB 1024x1024 OIG.jpg)

(167.36 KB 1024x1024 OIG.jpg)

(135.37 KB 1024x1024 OIG.Nz.jpg)

(195.67 KB 1024x1024 OIG.kXB.jpg)

(143.80 KB 1024x1024 OIG.p4.jpg)

(110.87 KB 1024x1024 OIG.S9EwGnACM.jpg)

(167.54 KB 1024x1024 OIG.jpg)

(187.86 KB 1024x1024 OIG.m.EM7RgkMBhwscf.jpg)

(173.37 KB 1024x1024 OIG.jpg)

(168.50 KB 1024x1024 OIG.jpg)

>>31928 early 90s aesthetic is actually really cute with abdl, or is it just because i was a baby in the early 90s
(133.44 KB 1024x1024 OIG.Zc9aB5tBt.jpg)

(87.17 KB 1024x1024 OIG.m6vw94qHeg.jpg)

(129.24 KB 1024x1024 OIG.Qfyjy.d39eR.jpg)

(179.23 KB 1024x1024 OIG.631nap6kcur.jpg)

(165.38 KB 1024x1024 OIG.N9BUFL2Nr.jpg)

(1.10 MB 1024x1024 00066-1649357823.png)

(1.11 MB 1024x1024 00082-2249596652.png)

(1.13 MB 1024x1024 00086-3921137735.png)

(1.14 MB 1024x1024 00083-2249596653.png)

(1.08 MB 1024x1024 00068-1649357825.png)

(241.64 KB 1024x1024 OIG.jpg)

(219.27 KB 1024x1024 OIG.FrFE90ROUaYBDm.jpg)

(233.19 KB 1024x1024 OIG.RA4Zu3quWJV.jpg)

(270.25 KB 1024x1024 OIG.jpg)

(171.44 KB 1024x1024 OIG.jpg)

(161.49 KB 1024x1024 OIG.jpg)

(169.14 KB 1024x1024 OIG..gGxy08CDqbfog.jpg)

(184.26 KB 1024x1024 OIG.jpg)

(158.07 KB 1024x1024 OIG.jpg)

(849.29 KB 640x1024 Aifog112923.png)

I'll animate it later
(287.73 KB 1024x1024 OIG.jpg)

(174.65 KB 1024x1024 OIG.HtGH0AkIX89yvZL.jpg)

(168.23 KB 1024x1024 OIG.4lDpLIUGGyY0u5pIn.jpg)

(245.63 KB 1024x1024 OIG.jpg)

which prompts are you guys using to getting such perfect diapers?
(682.46 KB 2880x2880 20231124_034443.jpg)

(684.40 KB 2880x2880 20231128_120114.jpg)

(633.96 KB 2880x2880 20231122_220519.jpg)

(731.83 KB 2880x2880 20231129_210019.jpg)

(718.05 KB 2880x2880 20231130_141513.jpg)

>>31511 Loved it. Especially the second and third image. >>31922 Amazing, the vintage polaroid look make it look even more realistic. >>31487 Second photo is pretty cute. >>31480 It looks like an indie film about abdl. >>31168 Japanese girls wearing diapers are the cutest thing in the world.
(166.81 KB 1024x1024 IMG_6934.jpeg)

(148.23 KB 1024x1024 IMG_6935.jpeg)

(151.47 KB 1024x1024 IMG_6936.jpeg)

A little bit of humiliation
(174.68 KB 1024x1024 IMG_6937.jpeg)

(186.51 KB 1024x1024 IMG_6938.jpeg)

>>31997 i usually type nappy with cute prints or puffy nappy with cute prints. typing the word diaper gives me a content violation.
(184.96 KB 1024x1024 IMG_6954.jpeg)

(189.49 KB 1024x1024 IMG_6953.jpeg)

(198.29 KB 1024x1024 IMG_6943.jpeg)

(145.65 KB 1024x1024 OIG.jpg)

(151.19 KB 1024x1024 OIG.jpg)

(159.57 KB 1024x1024 OIG.jpg)

(140.78 KB 1024x1024 OIG.jpg)

(138.80 KB 1024x1024 OIG.jpg)

(165.24 KB 1024x1024 OIG.jpg)

(167.93 KB 1024x1024 OIG.WQS55QT9ENlSPYo.jpg)

(138.56 KB 1024x1024 OIG.jpg)

(120.32 KB 1024x1024 OIG.Yvpc.jpg)

(135.91 KB 1024x1024 OIG.ZWOHXVgi3WyM.jpg)

(185.18 KB 1024x1024 OIG.jpg)

(180.80 KB 862x1024 OIG..jpg)

(164.28 KB 1024x1024 OIG.e_yS1poe8F.jpg)

(169.34 KB 1024x1024 OIG.jpg)

(137.99 KB 1024x1024 OIG.jpg)

(158.66 KB 1024x1024 OIG.p6nDnOqNqL7Chzw8Q_.jpg)

(148.77 KB 1024x1024 OIG.jpg)

(137.42 KB 1024x1024 OIG.P54b4DfXmeCFdKBsjs.jpg)

(168.47 KB 1024x1024 OIG.jpg)

(168.40 KB 1024x1024 IMG_6973.jpeg)

(194.24 KB 1024x1024 IMG_6982.jpeg)

(188.53 KB 1024x1024 IMG_6994.jpeg)

(182.18 KB 1024x1024 IMG_6990.jpeg)

>>31986 >>31987 >>31990 I really want to see the stinker win!
>>32050 God, those, especially the second photo, really look like those vintage ABDL photos from back in the 1980/1990s >>32150 >that first movie It's amazing to think that soon enough this tech will evolve enough to not only be able create a pixar abdl movie poster, but the whole movie. Crazy times, glad to be here.
are all of the tools you are using for this free?
>>32161 Yea, just sell your soul to the fap gods.
honestly, AI has changed the game. Who knew some of my best faps would come from AI generated art
(135.44 KB 1024x1024 OIG.jpg)

(130.48 KB 1024x1024 OIG.jpg)

(133.44 KB 1024x1024 OIG.jpg)

(129.33 KB 1024x1024 OIG.w.jpg)

(136.41 KB 1024x1024 OIG.jpg)

(146.65 KB 1024x1024 OIG.jpg)

(129.32 KB 1024x1024 OIG.jpg)

(146.59 KB 1024x1024 OIG.jpg)

(145.98 KB 1024x1024 OIG.skQfqb.jpg)

(140.92 KB 1024x1024 OIG..jpg)

(121.83 KB 1024x1024 OIG.jpg)

(130.09 KB 1024x1024 OIG.jpg)

(128.92 KB 1024x1024 OIG.jpg)

(137.02 KB 1024x1024 OIG.jpg)

(220.26 KB 1024x1024 IMG_7006.jpeg)

(151.10 KB 1024x1024 IMG_7010.jpeg)

(160.42 KB 1024x1024 IMG_7013.jpeg)

(133.47 KB 1024x1024 OIG.jpg)

(164.90 KB 1024x1024 OIG.jpg)

(146.86 KB 1024x1024 OIG.lxsj2Pu6JuSAKy8E.jpg)

(116.73 KB 1024x1024 OIG.EJi.jpg)

(139.32 KB 1024x1024 OIG.jpg)

(156.10 KB 1024x1024 OIG.1bD_NJXxEkn557Q.jpg)

(149.90 KB 1024x1024 OIG.jpg)

(155.01 KB 1024x1024 OIG.l5fb.9sn1qoRn.i.jpg)

(136.76 KB 1024x1024 OIG.jpg)

(149.49 KB 1024x1024 OIG.jpg)

(112.29 KB 1024x1024 OIG.jpg)

(129.63 KB 1024x1024 OIG.LGHQ3hOV69c.jpg)

(148.37 KB 1024x1024 OIG.jpg)

(101.59 KB 1024x1024 OIG.jpg)

(118.22 KB 1024x1024 OIG.y3AB698e.jpg)

(223.54 KB 1024x1024 IMG_7024.jpeg)

(205.45 KB 1024x1024 IMG_7022.jpeg)

Mom put Back in pampers
(161.38 KB 1024x1024 IMG_7027.jpeg)

(169.47 KB 1024x1024 IMG_7029.jpeg)

(152.96 KB 1024x1024 IMG_7030.jpeg)

Your new life
let me see what happens when i push to this to 60 seconds.
didn't come out as i had thought it would, anyways queue the next batch of dalle-3 images.
(1.28 MB 1024x1024 ComfyUI_00794_.png)

(1.21 MB 1024x1024 ComfyUI_00830_.png)

(1.28 MB 1024x1024 ComfyUI_00836_.png)

(1.79 MB 1024x1024 ComfyUI_00738_.png)

(1.28 MB 1024x1024 ComfyUI_00615_.png)

(252.66 KB 1024x1024 IMG_7039.jpeg)

(188.24 KB 1024x1024 IMG_7040.jpeg)

(189.27 KB 1024x1024 IMG_7041.jpeg)

A little bit more public humiliation
>>32271 what was the prompt from the first two
(1.06 MB 1024x1032 ComfyUI_01104_.png)

(1.08 MB 1024x1032 ComfyUI_01083_.png)

(1.71 MB 1024x1024 ComfyUI_00640_.png)

(1.05 MB 1024x1024 ComfyUI_00382_.png)

(1.04 MB 1024x1024 ComfyUI_00381_.png)

>>32277 Blonde 20 year old in a business blazar and diaper squats down as a group look at her surprised ((Pixar animation cartoon)) and Mature redhead squats down by toilet wearing a diaper ((Pixar animation cartoon)) Im using SDXL and several abdl loras I found on civitai
(167.78 KB 1024x1024 baby face 7.jpg)

(234.74 KB 1024x1024 baby time.jpg)

(115.10 KB 1024x1024 return to baby2.jpg)

the movie cover idea is so good I couldn't not do it
(182.21 KB 1024x1024 IMG_7059.jpeg)

(199.34 KB 1024x1024 IMG_7061.jpeg)

>>32282 wow wtf is this pieter bruegel the elder abdl content. More please, but with less hag.
I don’t know if this counts. But Google Bard’s ai writing system is surprisingly good at writing up ABDL scenarios. I’ve done a few ads featuring an ABDL Marin kitagawa and they have more or less turned out great in dreaming up my wank scenarios. Unlike chatGPT which from what I can tell simply bans writing fetish porn in general.
(859.81 KB 1216x768 stockphotos.png)

i really hate to say it but we really having a quality assurance problem in this thread. Maybe take a break from posting dalle3 images, make train some new ABDL lora's?
>>32304 Agree to disagree, the weird, surreal nightmare stuff is my favorite.
AI wizards please heed my request of Ariana Grande being punished for her persona of being a spoiled little girl with a spanking and diapers
>>32304 >this
(1.36 MB 1024x1024 ComfyUI_01882_.png)

(1.23 MB 1024x1024 ComfyUI_01852_.png)

(1.37 MB 1024x1024 ComfyUI_03798_.png)

(1.24 MB 1024x1024 ComfyUI_01940_.png)

(1.32 MB 1024x1024 ComfyUI_02076_.png)

>>32295 i used Hieronymus Bosch in the prompt to get that style. it's pretty prone to creating nightmare hags but here is one other from the batch.
Damn. This stuff is so advanced but every single time it's like the thumbnail looks good and my brain is filling in the details on what a cute scenario it is and then I expand the image and all the weirdness of it just turns me off. I wish it was possible to get more realistic looking diapers out of this stuff.
>>32305 I'm with you on this anon , it's like a glimpse into an another reality, dark fantasy and shit ,makes me rock hard
(3.17 MB 1536x1152 00259.png)

(1.50 MB 960x1280 00264.png)

(1.93 MB 1280x960 00278.png)

(1.53 MB 1152x1152 00280.png)

(1.91 MB 1152x1536 00283.png)

(1.94 MB 1816x1024 00054-3873520047.png)

Would be cool if there were some diaper+tights generation/models
It's just me or it's getting harder to get past the doggo?
>>32507 every single fucking day...
>>32508 Feared so. At least I got my fill
>>32503 provide the training material and I can make it happen
Hey AIFog, got anymore pics since your pixiv is still down?
>>32487 Who is this one based off of? She looks familiar...
>>32522 Gillian Anderson aka Dana Scully from the X-Files
do you guys pay for the AI? Or are there good free options?
>>32539 Pretty bunny. 😀
>>32538 The first one and last one have a weird Jonah Hill sort of face but are otherwise fantastic
>>32567 model?
>>32567 Those are decent. Cartoony but still good. Does it only work for fat women or could it do small and petite?
(1.43 MB 1024x1024 00004-3353914102.png)

(1.41 MB 1024x1024 00007-1476293591.png)

(1.50 MB 1024x1024 00005-1115374133.png)

(1.21 MB 1024x1024 00009-3295493121.png)

(1.30 MB 1024x1024 00006-4188677715.png)

(1.10 MB 896x1152 ComfyUI_09157_.png)

(1.54 MB 896x1152 ComfyUI_09482_.png)

(1.33 MB 1024x1024 ComfyUI_08419_.png)

(1.57 MB 1024x1024 ComfyUI_08540_.png)

(1.01 MB 896x1152 ComfyUI_09111_.png)

>>32582 Stable diffusion XL using a diaper lora >>32584 Its very flexible, it can do marge simpson
(3.74 MB 1664x2432 UpScaling171223_001.png)

(4.14 MB 1792x2304 UpScaling171223_005.png)

(5.03 MB 1664x2432 UpScaling171223_002.png)

(4.58 MB 1664x2432 UpScaling171223_004.png)

I been working on training my own LORAs. Getting the basic fundamental isn't hard. You don't even need that many pictures when training SDXL (I have managed with 10 good quality base images that I derived 15-20 additional images). But it seems like about 30-50 is the sweet spot. I tried with 50-100 and it seemed to lose details. The hardest part is getting texture and that god damn tape to not disappear. The tape has to be obviously clear and visible or else it just gets lots. Also avoid red features, it too easilly mistakes those for skin. I do just fine with 4060TI, I got plenty of VRAM to spare. I run patch size of 4-8 depending on how I prepped the dataset. I been getting consistent results with my training attempts atm, so I'm really just adjusting for getting more details. The diaper is supposed to be ABU Gator. For real now: Training is the easiest bit. It is mainly just waiting. Getting a decent dataset of even just 30 good images is a pain in the fuckin ass.
(325.61 KB 512x512 download-34.png)

(327.65 KB 512x512 download-36.png)

(325.17 KB 512x512 download-33.png)

(353.31 KB 512x512 download-37.png)

(356.16 KB 512x512 download-35.png)

(987.42 KB 1024x1024 00042-3767215768.png)

(1.18 MB 1024x1024 00028-4160389974.png)

>>32608 Looks good. Nice point on the dataset amount. What are you doing in terms of repeats and / or epochs? Are you using Kohya to train?
>>29801 Does anyone know if there is a Lora in which Japanese "omutsu" cloth nappies can be produced?
(355.27 KB 512x512 download-47.png)

>>32612 Yes Kohya_ss GUI to make it easier. For those I had 50 images (I purged down from 80 and got better results). LR is 0,0005; cosine; 10 repeats (This varies from 5-30 depending on the dataset - thus far just trial en error on my part); 11 epocs (save every 3rd), usually the 9th is good; fp16 (BF16 just makes a distort mess every time, no idea why); AdamW8bit. Thats about it really. I have debiased estimation on also, but works fine without also but makes it more consistent Beyond that I really don't play around with the settings. Probably should look into them more than just skimming through the papers and github things. But my experience thus dataset is 90% of the thing and as long as the training settings work - you get a decent LORA out of it. And I have done all sorts of funky things: Like train at very high LR 0,1 I shit you not, though it can NAN if you ain't careful. And as long as it works, it makes a working LORA. It is actually quite fucking amazing. I don't know why people make it out as some arcane difficult thing and with lots of urban myths without any basis citing back to some youtuber who doesn't know shit. In reality all you need to do is just do basic reading of the research papers and pull requests. I don't even know "how to code" and yet I can navigate that stuff - but I am an engineer of another totally unrelated field. The answers are in the documentation - if you are struggling then you should just read the documentation. As an example: there is this "FP16 and AdamW8bit doesn't work on SDXL! Use BF16 and adafactor" and that is just BS nonsense. They either had buggy bitsandbytes or bad setup otherwise which can cause obscure issues - but I have yet to find a configuration that doesn't produce atleast functional LORA as long as the training starts and doesn't make a NAN. But I can't make captions in dataset to work... I use them and I get either aggressive over fit or just nonsense. Maybe I need to train the text encoders also if I use captions? Dunno. Haven't had time to look in to it. I can do 2 runs a day after work if I am lucky. But if the GUI would allow queue of different training jobs I could leave it for when I go to work.
(1.23 MB 1024x1024 00111-1425962470.png)

(1.06 MB 1024x1024 00134-4254555060.png)

(1.10 MB 1024x1024 00147-263825951.png)

(988.45 KB 1024x1024 00174-427910008.png)

(1.19 MB 1024x1024 00163-580798932.png)

>>32627 Thanks for the insight. Here is the result of 60 images with captions, 5 repeats, 10 epochs and other settings most people would recommend on YouTube for SDXL.Going to try a bit more of what you suggested above also.
(1.07 MB 1024x1024 00244-2873265823.png)

(1.11 MB 1024x1024 00188-2027843178.png)

(1.16 MB 1024x1024 00202-3267691294.png)

(1.05 MB 1024x1024 00178-3782755811.png)

(939.19 KB 1024x1024 00182-2300334887.png)

Ok this one turned out better in details, and definition. Problem it can't be styled. Seems like trained with too many dimensions, which also picked up the definition of "photograph" in to it - making it impossible to force out of it. I did realise this early on that as few as 4 dimensions is enough to get things into it. It would appear that you got to find the lowest dimensions you can and run with that. This had 24, and I have had good luck with less than 16. So I might just try halving it and see where that takes me. I had 32 images, 10 repetitions, FP16, cosine, LR0,001, 10 epochs + 3 epochs with 0,0001 of which I took epoch 2 as the best. Getting the printed details in to the diaper is hard, because every picture had same elements but slightly differently. >>32608 Was easier to train to get the "look".
>>32631 I am curious to see AI photos on the hospital theme
I requested this from one of DeviantArt better anime-style ai artists. Since she has some diapered sailors and astronauts. I wanted a patriotic request. A diaper girl POTUS.
I'm enjoying the messing feeding and embarrassed pictures in strollers highchairs and such. Please keep posting more of these. Which AI image maker are you using and what input strings are you using for these?
(98.64 KB 1024x1024 OIG.kXtTjO2xSL7Er9.jpg)

(127.45 KB 1024x1024 OIG.BPzHp6AyCTn.jpg)

(107.77 KB 1024x1024 OIG.nsKIGkI8hu0.jpg)

(104.61 KB 1024x1024 OIG.HaXfn_wCjf6.jpg)

(136.38 KB 1024x1024 OIG.8rXQgD4meesZv.jpg)

(169.82 KB 1024x1024 OIG.sJNj6iB4vWyPzCk2.jpg)

(146.15 KB 1024x1024 OIG.F5LwHEu5QQ.jpg)

(130.73 KB 1024x1024 OIG.tPTNgu60oFV.jpg)

(130.95 KB 1024x1024 OIG.uTi3vuAe.jpg)

(134.25 KB 1024x1024 OIG._I9eTScw6LxM.jpg)

(137.57 KB 1024x1024 OIG.lfnQgwV59hHSf4.jpg)

(177.67 KB 1024x1024 OIG.ywWEZ6ts6.jpg)

(136.16 KB 1024x1024 OIG.5N8b7uAzg.hFNd.jpg)

(154.26 KB 1024x1024 OIG.UthuZJsaPTMM_u.jpg)

(166.67 KB 1024x1024 OIG.bghEgXcYyNB.jpg)

(142.52 KB 1024x1024 OIG.GaUCqw8q3kFC4o.jpg)

(212.62 KB 1024x1024 OIG.Etx29KCizl.jpg)

(124.32 KB 1024x1024 OIG.CZD6FNsL_jkFpP..jpg)

(170.24 KB 1024x1024 OIG.d9Pj9yvG80.jpg)

(157.28 KB 1024x1024 OIG.X9VUYnDX2.jpg)

>>32631 This is amazing. Can you post / upload the model? (±Creampie diaper pics?)
(245.74 KB 1024x1024 IMG_7496.jpeg)

(207.76 KB 1024x1024 IMG_7146.jpeg)

(184.13 KB 1024x1024 IMG_7137.jpeg)

(216.23 KB 1024x1024 IMG_7136.jpeg)

>>33095 Love the Pampers store in the background. Walk of shame there for any bratty girls.
woo are you using bing?? It's so hard to make him better in diapers
(367.97 KB 384x640 00098-0.png)

(376.82 KB 384x640 00220-4.png)

(392.53 KB 384x640 00191-7.png)

(415.36 KB 384x640 00237-5.png)

(766.67 KB 512x800 00553-0.png)

Here's some of my good generated images. I'm still a beginner, but I'm getting better at this.
>>33096 What description did you use for these?
(226.49 KB 734x834 article03.jpg)

(714.62 KB 1000x765 Screenshot 2024-01-06 103529.png)

(234.42 KB 734x831 article02.jpg)

(1.39 MB 1024x1024 00025-1176817777.png)

(1.19 MB 1024x1024 00001-4138387451.png)

>>32898 Sure here is the Lora model for SDXL: https://mega.nz/folder/QmlQlDjY#Y29cAMMjlEFeYzUF0xeV2Q There are 3 different models that were all saved at a different epoch, I use 000007 the most. Far from perfect but decent for my particular tastes.
>>33155 What model was it trained on, what trigger words did you use, and what weights will give us the best results?
>>33155 Loving the Aubrey stuff
(1000.38 KB 1024x1024 00070-1661790225.png)

(1.01 MB 1024x1024 00004-378522267.png)

(1.17 MB 1024x1024 00009-2523045005.png)

(1.24 MB 1024x1024 00020-743410467.png)

(1.01 MB 1024x1024 00029-2380600755.png)

>>33155 >>33161 Trained on SDXL base - sd_xl_base_1.0.safetensors Trigger word is "abena diaper" (you can also try "abenagirl woman") All of these were at :1, I haven't tried any other weights. I'm new to all of this but I like how the Lora turned out.
(1.25 MB 1024x1024 00040-1917259969.png)

(1.10 MB 1024x1024 00046-2384339082.png)

(1.06 MB 1024x1024 00050-4145191974.png)

(1.34 MB 1024x1024 00067-2519813875.png)

(1.34 MB 1024x1024 00082-2893106741.png)

>>33155 Those are great. I would love to see some like that with Margot Robbie, she seems like a perfect fit for diapers.
>>33181 I wasn't quite able to get as good faces as you without inpainting, but otherwise it's a pretty good lora. It does tend to add a watermark though so maybe consider cropping your pics if you make another.
https://aidiapers.com/ Is this legit? It looks scammy. >>33186 I'm not calling you out or anything, but I do wonder why you think Margot Robbie would be a perfect fit for diapers.
>>33201 The idea that people will pay to generate fake porn, but refuse to pay content producers of real porn just astounds me
>>33204 I understand that. I wouldn't mind paying for porn and ABDL models more if it was cheaper and better. However, even thouh there's some really cute models out there, and there's a few companies out there that are willing to do custom (within limits) scenarios for you if you commission them, unfortunately most of the models out there who are willing to do ABDL at all, aren't that attractive. And even if they are attractive, the vast majority of content getting produced for this fetish is literally just women alone in their rooms pissing themselves with no dialogue and without even cracking a smile. There is some variety out there and there are some examples of some good quality content out there, however, paysites with actual curated sets (whether it's nursery or not) and also just content with diverse, interesting scenarios, costumes, and other models, is so much harder to come by nowadays. So yeah even if AI art is kinda bad and even if I'm willing to pay for real porn I kinda want to see more AI diaper porn getting made because I hope that eventually, once the technology progresses enough, we'll get to the point where it's actually able to produce the sort of ABDL porn I've always wanted to see but only barely been able to get scraps of with real women producing it.
>>33204 >The idea that people will pay to generate fake porn, but refuse to pay content producers of real porn just astounds me You mean people would rather pay for interesting content that AI can provide rather than the 3000th generic camwhore wearing whatever diaper is trendy and providing nothing other than having tits?
>>33201 It's not very good. Tried an example and it gave a messed up face. Probably uses something like Draw Things, which you can run on your end for free.
(181.37 KB 1024x1024 IMG_7543.jpeg)

(216.56 KB 1024x1024 IMG_7540.jpeg)

(272.44 KB 1024x1024 IMG_7501.jpeg)

(309.03 KB 1024x1024 IMG_7565.jpeg)

(248.07 KB 1024x1024 IMG_7574.jpeg)

(251.90 KB 1024x1024 IMG_7577.jpeg)

(283.36 KB 1024x1024 IMG_7542.jpeg)

Sent back to the nursery 🍼
(374.35 KB 384x640 00027-0.png)

(586.78 KB 960x576 00057-2.png)

Couple more of my good results. Pacifiers and the Jack-o pose were really tough to get right.
Managed to get exposed tits on one without even asking for it. Method has been similar to other anons here, just ask for what you want in the context of a medium like a movie ad at a mall or something similar. >>33250 >>33249 >>33247 Prompts for these?
>>33155 Got anymore of Aubrey?
(242.52 KB 1024x1024 IMG_7644.jpeg)

(303.28 KB 1024x1024 IMG_7650.jpeg)

Babys day out
(139.20 KB 1024x1024 000003.jpg)

(160.67 KB 1024x1024 OIG.RpkPCuDWS.jpg)

(145.36 KB 1024x1024 000002.jpg)

(145.28 KB 1024x1024 OIG.d2O0YHLbQ_btVw_0.jpg)

(187.35 KB 1024x1024 000001.jpg)

New pandemic diaper mandates just dropped. No entry permitted into public businesses or places without one.
(208.26 KB 1024x1024 OIG.Wo.jpg)

(170.05 KB 1024x1024 OIG.jpg)

(174.40 KB 1024x1024 OIG.jpg)

(219.53 KB 1024x1024 OIG.jpg)

(153.15 KB 1024x1024 OIG.jpg)

OK; I got Comfi UI installed, SLDX and the refiner, as well as some of the LORA's posted here, but most of the time I'm still getting nightmare fuel. Anyone willing to share their negative prompts, incase I am missing something obvious?
(1.06 MB 1024x1024 00051-2217465853.png)

(1.05 MB 1024x1024 00047-589075039.png)

(1.06 MB 1024x1024 00056-611204097.png)

(1.08 MB 1024x1024 00057-877054001.png)

(1.21 MB 1024x1024 00042-3263920335.png)

>>33506 Not too familiar with ComfyUI but give us more info on what you're doing. Prompts, settings, etc. I use Automatic1111 and I used this model for all of the M4 model images I uploaded here including the attached images. Its a much better model imo, but CivitAI has plenty: https://civitai.com/models/140737/albedobase-xl
(356.84 KB 1510x898 workflow.png)

>>33512 I'm totally new to all this, just started learning two days ago, so I apologize if I come across as totally clueless, cuz I am. I was recomended to use a user made pre-configed workflow, as supposedly the settings are already mostly all plugged in, and I should only need to adjust and tweak a few things. What I did was add in, say, the AbenaM4 Lora posted earlier between the base model and the refiner, as shown in the image, then I entered a prompt; "A photograph of a blonde woman with pigtails and blue eyes, wearing a pink shirt and wet abena diaper. She is sitting in an adult sized play pen, a big teddy bear on her right. Her shirt says "I <3 DADDY" on it." It says it allows for supporting terms in the other entry box, so in there I have added "diaper, abena diaper, stroller, blonde hair, teddy bears, rattles, pacifier, baby bottle, infantile, baby goods, wet diaper" As for negative prompts, I am using a mix of stuff I found online that seemed appropriate; "noise, grit, dull, washed out, low contrast, blurry, deep-fried, hazy, malformed, warped, deformed,bad anatomy, bad hands, missing arms, extra hands, extra fingers, bad fingers, extra legs, missing legs, poorly drawn face, fused face, worst feet, extra feet, fused feet, fused thigh, extra thigh, worst thigh, missing fingers, long fingers, extra eyes, huge eyes, amputation, cartoon, cg, 3d, unreal, animate, cropped, blurry". Lemme know what else you may need to know; again I am totally new to all this, and figuring shit out is fine, but getting tips from those who seem to have things figured out will save me some time I figure.
>>33535 Hm, your workflow is a lot busier than mine. However one issue I get is the refiner will 100% of the time stop the diaper from looking like a diaper.
>>32608 could you post your diaper LORAs please? your AlphaGatorz look incredible, we probably need multiple LORAs for diff diaper types
(1.58 MB 1024x1024 00032-1874666734.png)

(1.31 MB 1024x1024 00044-2067056311.png)

(1.35 MB 1024x1024 00041-1163807501.png)

(1.48 MB 1024x1024 00026-564307165.png)

(1.30 MB 1024x1024 00046-971125940.png)

A quick cushies lora I threw together. I'm surprised I was able to get half decent generations with only like 25 training images.
(1.39 MB 1024x1024 00003-1449656978.png)

(1.39 MB 1024x1024 00016-1666266252.png)

(1.42 MB 1024x1024 00012-950060768.png)

(1.42 MB 1024x1024 00005-3045677209.png)

(1.43 MB 1024x1024 00004-94710746.png)

>>33881 VAE was fucked in those. Here's some better pics.
>>33881 >>33905 Interesting. Not bad for only 25 training images. What model did you train on? Standard SDXL or something else?
(290.00 KB 1024x1024 IMG_7831.jpeg)

Embarrassing Diaper checks by McDonald’s worker
What about some Taylor Swift diaper porn? I found many of the porn did with her to be too hardcore. It would be nice seeing some cute AI images of her having her diaper change as a cute little baby.
>>33953 Yep just standard sdxl.
>>33971 Wasn't there something that went viral about her fans wearing diapers on tour?
(1.61 MB 1024x1024 00009-2749092485.png)

(1.40 MB 1024x1024 00016-942956071.png)

(1.39 MB 1024x1024 00021-197143123.png)

>>34013 I might upload my cushies lora to civit, but I don't plan on it.
(1.48 MB 1024x1024 00036-2710902364.png)

(1.37 MB 1024x1024 00012-3757422348.png)

(1.35 MB 1024x1024 00017-1674999129.png)

(1.30 MB 1024x1024 00043-5761363.png)

(1.37 MB 1024x1024 00011-2930198081.png)

>>34018 Loras are getting much easier to create. Here's a barebum lora to go along with the cushies lora.
What are the prompts to get it as an actual diaper and not just panties? I've almost got a good one. Trying to have an 80's anime style Japanese woman wearing a wet diaper with animal prints on the front.
>>34126 these are great! pls upload and post if you're able :3
(1.38 MB 1024x1024 00060-1511233077.png)

(1.45 MB 1024x1024 00011-2936709350.png)

(1.51 MB 1024x1024 00042-3785972325.png)

(1.38 MB 1024x1024 00024-302797078.png)

(1.52 MB 1024x1024 00051-844921255.png)

>>34221 I want to retrain with the faces all inpainted before I think of releasing it. I could share my Kohya json settings if anyone wants though.
>>34190 Thanks! Loved the first one.
Can I do these without a potent pc? My computer is modern but some of the specs for Ai generation are crazy, can I do it online somehow?
>>34294 If you have a PC you need 8gb of vram or higher and a good diaper lora to use with stable diffusion. I think you can use bing online, but I personally think the model it uses is trash.
>>32627 I have the same issue with non sense and overfit when training dataset. I tried to use caption files with an activation word and also with descriptive words. Are you telling me that I would get better result by just training without any captions? How are you activating the Lora without an activation word afterward?
>>34357 The class prompt is the activation word, it gets trained regardless whether you use captions or not. Instance and Class tokens get trained. I recommend keeping Instance "nonsense" so it remains pure. Unless you use a LORA type which allows training of many concepts, then consult the specific instructions for those.
>>34358 Thanks anon,I’ll to train without any caption and I’ll see if it’s better.
>>34357 The captions for my barebum lora were all along the lines of "a woman, wearing a barebum diaper, sitting on a bed, pink kneehigh socks" without the quotes. I just made sure to repeat "A woman, wearing a barebum diaper" In all of my captions. I swapped things up if it was a man. You could probably get away with saying "A person" as well. I think my lora came out pretty well.
>>32608 I trained the ABU Gator again was able to get it to perform better. However it still isn't as good as I want it to. Seems like big resolution dataset isn't a best. I get better resoluts with varying resolution with a big range. >>34361 This is not an universal rule. You have to test out case by case specific. I have had most success with eliminating as much stuff as I possibly can and making the training extremely focused and dampened. For the longest time I trained ONLY with AdamW8bit. I got extremely good results. Now I managed to get DAdaptAdam to work - which is what I used to trained this. However I recommend actually reading the papers behind the tech, they help you alot. Along with watching few long form technical youtube videos. Avoid anything with clickbait title or "Automatic1111" or such in the titles, as they are fishing for views. They have no answers. Any video saying that they got ready answers are wrong. Because if they had ready answers they'd be earning millions in the tech space and not making youtube videos about it. This is very deeply technical stuff if you want to do it well. I don't want to discourage you, even though it might sound like it. You can do surface level stuff and dip your toes and to LOTS of stuff with basic copy paste settings. Hell that is how I started. But if you really want to do well, you will need to understand what it is your are doing. You don't need to be a coder, or software person at all. The fundamental math and logic behind all this is detached from that. Oh and here is a thought. I have concluded that the token "diaper" is polluted beyond reason. Avoid training with it. Come up with just about any other token for underwear or such. I often add diaper to negative tokens when generating. Example. All the diapers I trained, I trained as "underwear", "underpants", or such tokens. Get out dictionary and find a foreign word if you have to. But I have only been able to get anything worth a damn on "normal" models. I can't get ANYTHING done on Booru tag models. Absolutely nothing of use. And yet another thing. All models of the same type can be used to train LORA for that similar types. As in all SDXL models work for training of LORAs for SDXL models. By my experience merged models are worst, so stick to trained models.You need test around a bit. My goto currently is AtriumV2, and the worst are Juggernaut 8 and 7.
>>34363 This is helpful,and you’re not discouraging at all. In fact,I love all the tech behind and am a « bit » knowledgable in it since I have a machine learning degree from six years ago,but stuff is evolving so fast that my knowledge is constantly challenged. I guess I’ll take it more seriously and read all the doc about the tech I missed in theses last years.
>>34366 Right. So you knowledge and skillset should be greatly superior to me. I'm myself manufacturing engineer, I took 1 module in basics of coding with C/C+ and Python which I managed to pass - lets leave it at that - as a mandatory elective as part of my degree and this was early on. I realised then and there that I do not have what it takes to do software side things. I'll take a mechanical system any day over code. But the fact is that my degree gave me the relevant skills set to read those papers. Even though the maths involves is way beyond me. But I can figure out enough from those papers to approximate what it is I need to do and why. Once you realise the "thing" you get this sort of enlightement and things start to make sense. You start to see the system you are manipulating. Once you can visualise it, everything after that is just details. I have no specific idea about how this training works at individual component level. But I sure as hell got a idea about what kinds of components there are under the hood. I'm sure if I had more time and energy (along with hardware, I push on with a HP prebuilt into which I crammed the only 4060TI that I could fit into it and run with the puny PSU it had for that 16gb of Vram. (You don't need 16 GB, 9 GB is enough for SDXL - for those that are interested and reading this chain. 13 GB is about the minimum for full quality without having to compromise in speed)). And honestly... I have mostly just experimented. Just done stuff. You start to see clear patterns of behavior very quickly by just doing stuff. Don't overthink the process. Just do it. You are better off doing failed attempts than no attempts at all. Also remember: You only need about 10-15 pictures which are good quality. Then you can with very basic photoshop and upscaling trickery turn these to 5-15 additional pictures. And keep in mind that you don't need to have the photoshop good enough to fool people, you just need to fool the AI... The AI is very stupid, it only knows the average of the thing you are training. High quality dataset beats quantity instantly. If you get an alright LORA trained, you can generate additional pictures to train with. I did a whole style TI, Dreambooth (before we had LORAs) and LORA like this. I made a version, then generated ton of pictures and chose the best for the next iteration. You just have to remember to take them to photoshop and add noise, adjust the image quality and give em a blast of JPG compression and you good to go.
>>34363 Good info, figured that common words are tainted with poor quality images. Did you mention how many steps and batch size before, how long does it take to train?
>>34374 Batch size really doesn't matter, but you should keep it as high as you can with the vram you got - this is the general advice for all training. You can also use gradient accumulation. Accumulation of 4 with batch size of 1 is the same as batch size 4 with accumulation of 1. Personally I haven't felt it affect anything really - it feels to find a solution quicker but it is still about the same. There is no "right answer" to this. Only thing that is true is that higher batch size is more efficient. Steps? Well.. I don't know. It depends.I use DAdaptAdam a lot atm, and it adjusts LR by itself. I allow the script find the max steps by itself (Which is generally 5000 or so. So with my batch size of 4 it is 1250) this takes 1½ hours. It varies a lot since it depends on the resolution of the images in the dataset. I'm not going tell you a number here, because I don't have one. But I well tell you this: DAdaptAdam = More epoch than repeats. It also sets LR to basically 0 when it has found a solution after nothing will happen. AdamW = More repeats than epochs. I haven't managed to get Prodigy to make good quality things. All the optimisers have their own features and downsides. However keep in mind that this is just for normal LORA. The Lycoris and such, I haven't touched yet. Look... Don't take my words as gospel, just experiment. I don't know what and how you are training. I don't know your base model. I don't know your dataset. I have no idea about what it is you want to make. Just experiment, that is how I figured this shit out. Good quality dataset gets the solution quicker. I have managed to improve training and use less steps by culling the dataset of bad examples. How do I know what are bad? After you got the LORA you can test it and you see features and elements which are "wrong", find those in the dataset. You can also amplify the LORA when generating to find these oddities. But I have noticed that patterns on things other than what I want to train mess up training: Like when I did the Cushies >>32656 LORA. I learned that pictures where bed sheets had patterns, so these started to appear on bedsheets when generating. Solution to this was to remove the pictures or photoshop the bed sheets to not have pattern. It would appear that the LORA "understands" that the subject had a pattern in it and learned the pattern; but if dataset had other patterns it it learned that "patterns exist elsewhere... So add this pattern to everything else ALSO". Also another note. If you got red/orange tones in the subject and it is like piece of clothing (like a diaper) then during generation there is a high change it will blend in to skin. No idea how to deal with this other than adjust the colours on the subject in the dataset. Seriously... Go experiment. Don't bother trying the get the settings perfect instantly. But if you manage to overfit the thing (or parts of it), then you know that you have too many steps or too high LR. So you are within the correct postal code. If you get absolute nonsense mess, then you are not.
>>34444 Something about these pictures is way too depressing, almost like they’re for a funeral.
I like this first one. The one thing I would change would be replaced her cup with a baby bottle, and make her dress shorter so that we can see her diaper. Sadly, whenever I tried to use the words baby bottle or just bottle, the algorithm says it hates me.
I honestly can't wait until this technology gets good enough to generate whole movies.
>>34463 What Lora / model? Looks great!
>>34465 Love these, the dresses with a diaper peeking out are my weakness.
>>34492 It's a custom model on top of just regular old sdxl using ReActor for faces.
>>29801 I don't know if this can contribute to this thread but Stable Cascade is out, a model that achieves higher quality than SDXL with lower inference times here the link on hugging faces https://huggingface.co/stabilityai/stable-cascade
>>34512 I was always paying attention to this earlier and, honestly? It was somewhat disappointing. It's no where close, not even to MJ 5 or so. And from what I tested, it still can't produce women wearing diapers out of the box :\ If you already have a model based on SDXL, there's probably no utility on training again on Stable Cascade. For anyone wanting to test out: https://huggingface.co/spaces/multimodalart/stable-cascade
“I now pronounce you father and daughter. You may change her diaper.”
A few more. The third one the body is totally screwed but I'm posting it nonetheless. The 4 forth wasn't quite the diaper I wanted it to be, but I really enjoyed the composition.
>>34542 Can we cool it with these bing images? They all look the goddamn same.
>>34545 Theres no words for how shit these are
>>34553 This is the equivalent of pasties and shitty paint "edits". Just use your daily allowance on some shitty generator site and spam it everywhere as if it is the greatest shit. Don't put any effort or thought in to it. Those are what? Bing generated? >>34363 Anyways. I got the Lora to be more versatile and reliable. Still cocks up some of the finer details, but it is like 60% reliable. However I'm better at making LoRAs than generating in my opinion.
AI is sad and retarded and you should kill yourself if this is the only way you can get a woman to engage with you
>>34655 these are looking good! thank you for sharing.
Tried my hand at a messy LORA. You can tell I used too many images from behind, but it's still half decent. The hardest part about LORAs is finding decent pictures. Too many terrible crotch shots exist.
(1.46 MB 1024x1024 base_output_00061_.png)

(1.51 MB 1024x1024 base_output_00060_.png)

(1.54 MB 1024x1024 base_output_00057_.png)

(1.30 MB 1024x1024 base_output_00082_.png)

(1.32 MB 1024x1024 base_output_00083_.png)

Best that I've gotten
>>34553 You cannot always tell if an image was made by ai, but you can always tell if there was an artist behind it. Now, for the price of one diapered maid, I will give you another cheesy line.
>>29880 wunderschön,so kann mann auch den urin einfach laufen lassen ohne ein klo aufzusuchen
>>34887 I'm a big fan of maids in diapers.
If anime is your thing, the VivaldiXL checkpoint on civit does some pretty good diapers. I'm not into weeb shit but you could probably get better outputs with better prompts. Even so these are pretty good.
>>34834 Damn, I'd love to see more like these. I've never managed to gen diapers under clothes like this. It always blends the diaper and pants together when I've tried.
now that the text to video ai is here, i'm curious to how that would work with porn. I mean, aside from the correctly illegal cp generation (fuck these guys), wouldn't ai text to video create an overwhelming amount of niche fetish porn? There is a demand, and computer nerds with weird fetishes will figure out quite quickly how to do this....
>>34990 Well we have had text to video for a while, just very limited since it went to about 5 seconds (Because adding more dimensions increases model size and computational time). But text2video really isn't anything else than text2img just done many times. The greatest problem is consitency from frame to frame, which is what the models are able to do, since they are taught that a thing can persist for longer. Honestly I think most of the generated content will just be the kind of aggressive kink spam we have seen here and other places. 0 effort and every result gets posted as the best thing ever. When it comes to csam material... well... if history has proven anything then you can't fucking do anything about it. There been cases in the news where pedos have gone back to swapping physical media and even tapes. I kinda hope that these AI models can cut the demand side of abuse material, reducing the criminal networks incentives for human trafficking and generally the financial interest in it. Maybe we can save lives like this - which is already more than what seem to be achieved atm.
>>34897 Thanks! This is really good.
>>35013 Mind helping me out with simple instructions for getting this working? I looked the other day but it looked real overwhelming.
(1019.69 KB 1024x1024 00081-698521452.png)

(1.02 MB 1024x1024 00096-763546582.png)

(1.07 MB 1024x1024 00185-3028796062.png)

(991.27 KB 1024x1024 00238-1831521280.png)

(1.09 MB 1024x1024 00268-3416843019.png)

>>34897 >>35013 Agreed. Very pleased with the outcomes so far.
>>35046 Can you share some of these prompts? Amazing
>>35040 Sure, Here's my prompt for this image >>35013 really i just reworked an example prompt from Civit-ai I had my own custom lora in there but i really think it's has 0 affect so it shouldn't be needed score_9, score_8_up, score_7_up, score_6_up, high quality, masterpiece, 8k, highres, detailed, , 1girl, ;O, black_thighhighs, blunt_bangs, bow, colored_skin, extra_eyes, fangs, fewer_digits, diaper:1.3, full_body, sitting in field, hair_bow, looking_at_viewer, one_eye_closed, drooling, pink_hair, pink_shirt, shirt, short_bangs, short_hair, short_sleeves, short_twintails, detailed_background, skin_fangs, smile, solo, diaper_girl, thighhighs, twintails, spread_leg,<lora:DDiaper-SDXL-Aifog-v37:1>, ddiaper, young-adult, adult, ((mature)), heart iris, loving, blushing Negative prompt: 3d, cgi, sketch, worst quality, low quality, deformed, blurry, realistic, censored, bad bad anatomy, hands, fingers,anime, Steps: 30, Sampler: Euler a, CFG scale: 11.5, Seed: 199407257, Size: 768x1024, Model: vivaldixlOpenDalle3_v10, VAE hash: 63aeecb90f, VAE: sdxl_vae.safetensors, Clip skip: 2, Token merging ratio: 0.3, RNG: CPU, Lora hashes: "DDiaper-SDXL-Aifog-v37: 7c8b8a9392c0", Discard penultimate sigma: True, Schedule type: karras, SGM noise multiplier: True, Eta: 0.12, Pad conds: True, Version: v1.7.0
aifog, maybe post your loras somewhere? there are some that have been up on huggingface for months now and not taken down
(1.22 MB 800x1200 Normal_CheckLora_00419_.png)

(1.19 MB 800x1200 Normal_CheckLora_00425_.png)

(1.16 MB 800x1200 Normal_CheckLora_00433_.png)

(1.17 MB 800x1200 Normal_CheckLora_00439_.png)

Any reason there aren't any new loras on civit? Lack of people working on them or are they getting removed? Still working on getting the tapes to look better, but progress is being made
has anyone tried yet the video generating AI tools for creating abdl scenes ?
>>35049 I can't wrap my head around booru models... The whole tagging system is just a language I can't handle. >>35077 My loras: >>34655 >>34363 I haven't posted anywhere. I'm not sure if I can be fucked with Civit - that seems like a drama zone. And I can't be fucked to register accounts. Anyways. I been working on Goodnites atm. My issue is that I'm shit at prompting. My issue is that I'm not sure if the Lora is bad, the base models isn't good OR that my prompts are bad. And I can't stop it from forcing in beds and bedrooms which always just look wrong on so many levels. Also I struggle to find decent images for datasets. ABDL community is really not that imaginative with pictures. Just close-up crotch shots. I need atleast 4th of the dataset to be... not a close up to prevent closeup bias.
>>35084 Where'd you make these?
>>35085 On my computer. I don't use any online services or paid systems. I train my own Loras because a RTX4060Ti is more than enough for SDXL lora training. Once I got some cash planning to get a CUDA card that could possibly natively train whole models, though I'd need to upgrade the whole computer for that.
is gemini worth a damn?
>>35062 most of them aren't really good, i want to make a set for pony based SDXL models
>>34706 These are insane. Someday I hope the technology that created these images reaches the idiot-proof consumer level
>>35136 I mean like it is frankly quite idiot proof already. Something like Fooocus. It just download the thing: click to run and it does everything including downloading of the model. And there been quite few "step-by-step" Dreambooth google colab notebooks - I'm sure they gotten better and include LoRA. I haven't checked for a long while. The only reason these haven't reached the level of Windows .exe programs is that things move too fast to develop and keep those updated. Also... No money in that, the money is in cloud services.
>>35208 Kohya_ss runs perfectly fine on windows with my 3070 and there is a one click installer for it. I have no idea what every option does and just use a json that someone online posted and it just seems to work.
Fooocus is 100% of the way to get started. It can run on a potato (GTX 1070) albeit at about 0:45-1:30 min per image. One click download and it just works. It has all the settings so you aren't limited if you learn it and the defaults improve the prompt to get good results right out of the gate. You won't be making anyone wearing a diaper without a lora though.
>>35232 I use Kohya_ss on a 4060TI, the 16gb allows me to do just about everything par for native finetuning of a whole model. I run batch size of 9 with FP8 with pictures ranging from 1024-1340 in resolution. And I got a fairly good idea about what nearly all the setting do - which is why I know that most of the are pointless to interact with. If you don't have +16gb of VRAM you just turn on all the memory saving options and thats about it. The interface and UI/UX is shit - but this holds true to just about any software nowadays. This is what makes it seem intimidating. In reality it is not, and the hardest part is explaning to people why they shouldn't be intimidated by it. And I have done this to people. And I have read the papers and technical documentation on most of the things that are relevant - issue is that I can't really understand the code that well so the actual code files are nothing but "This thing links to that..." and the fractal fuckery of these repos is beyond my comprehension. But having the basic understanding about what is going on under the hood helps - and that really isn't complicated, you don't even need to know the math for that. After a while of trying you start to see patterns in the way things train, and you get intuition about your dataset and settings. I personally write my thoughts to a small notebook on my desk. Then I just click "train" and occasionally check that the loss hasn't gone to NaN and get back to my watercolour painting while I wait. Just as with most technical and technology things. The actual thing you are doing isn't that hard in reality. The hardest thing is to get over the insecurity. I still say that I'm way better at making LoRAs Textual Inversions and the like, than I am in prompting. Which is why I have no fucking idea how boorutag models work - and therefor I don't use them. My point is: This isn't hard; the hardest bit is getting over the intimidating first steps.
(1.25 MB 1024x1024 Normal_W10CFG7_00202_.png)

(1.40 MB 1024x1024 Normal_W10CFG7_00003_.png)

(1.31 MB 1024x1024 Normal_W10CFG7_00029_.png)

(1.29 MB 1024x1024 Normal_W10CFG7_00209_.png)

Made progress. I'm happy with this one, but I think I can train it a bit longer. It fucks up faces a bit though. Not sure how to fix that.
(1.20 MB 1024x1024 Normal_W10CFG7_00183_.png)

(898.49 KB 1024x1024 Normal_W10CFG7_00278_.png)

(1.34 MB 1024x1024 Normal_W10CFG7_00044_.png)

(1.15 MB 1024x1024 Normal_W10CFG7_00019_.png)

>>35244 I tend to just not care about faces and either inpaint them after the fact or use something like adetailer to auto inpaint the face. Great images either way though. Is your lora creating all 4 tabs consistently or are these just the best of the generations? For megamax style I can never seem to get my tabs to be consistent. 1 in 5 end up looking acceptable.
(6.77 MB 1280x4096 CopaxTimelessCFG7_00001_.png)

(7.08 MB 1280x4096 PonyCFG7_00001_.png)

(6.69 MB 1280x4096 PixelCFG7_00001_.png)

(6.51 MB 1280x4096 JuggernautCFG4_00001_.png)

>>35256 Thanks. It gets the tapes right more often than not. Maybe 3 or 4 out of 5. The CFG and the checkpoint really drive it though. I dunno if these grids are too small after upload, but here's the full set for CopaxTimeless, Pony, Pixel, and Juggernaut. I have no idea how to prompt Pony, but it seemed to already understand diapers. The lora forced the white diaper blue tapes style when it worked.
(3.61 KB kohya.json)

>>35281 After looking through all that data I can see where you got some of your training data. Would you mind uploading the json file you use for kohoya so I can see how close your settings are to mine? Here is mine for anyone who wants to try it out. I typically try for 10 trainings of each image and a batch size of 10.
>>35288 Sure. We are pretty different in terms of settings. I gave up on Adafactor and Constant. Dadaption Cosine was what I ended up on. Prodigy was alright too. The 15 epochs wasn't really any different than the 10 from the model prior, but I added 2 more ass shots to correct an issue there and wanted to see if I could overtrain, but Dadaption doesn't really do that. I used 32 full body images after the ass shot addition. Took 12 of them and cropped them to close up diaper shots, and did a second copy of the close up with the background removed. 7, 1, & 2 repeats respectively. WD tagger. I tried natural language captions from CogVLM, but they didn't give as good results. As an added bonus, here are some more furries since I figured out Pony a little better today.
>>35291 I like the baby lion. :3
>>35288 How are the other LoRA types? I been just using the normal original vanilla LoRA. I haven't really bothered to look in to the other variants - as I think I still got a plenty to master just on the regular. There are some actually really interesting kinds available there for sure with really cool properties. Oh. And I recommend trying other base models (Which been trained instead of merged). I have solved many problems I come across by using a different model. Those LoRAs also work in all base models - only struggle with the merge extravaganzas which barely perform on general level to begin with. However if it doesn't train on the SDXL base model, it ain't training like that at all and you need to rethink. >>35291 Adafactor constant is not that good really. Really hard to adjust. Adafactor seems to only work with Adafactor scheduler. I personally use cosine for all and it works well. I use DadaptAdam with Cosine and it always works. Might not be the best, but the one I know will work every time. Most difficult part is setting the repeats to just right amounts. It seems to be around 7-10; 6 if the thing is obvious but below that it is just nothing. Above 10 it becomes inflexible. But if you struggle with bias to style, like it making only photographic, only anime, only painted like... etc. Throw in few random images of different styles. Doesn't need to relate to the subject itself. Like if I want to train something with only photographs, I put in a random cartoon character and it adds a lot of flexibility. And so forth. Also you can increase image quality (if your model is learning "low quality" attributes) but putting in few really good images of unrelated matter. Same thing with fixations. You can put in like partially exposed diaper, and you get extra flexibilty. But treat this like spice, just a small dash is enough.
>>35306 I can't say I notice a whole lot of difference training a locon. I had read some guides online that said they were better for lower end cards. I only have 8gbs of ram and have ran into memory constraints before. I may try retraining a model using a different model trained on sdxl just to see how it performs vs a standard sdxl model. Just trying a lora with dadaptation caused my training time to spike to over 48 hours. I'm guessing my 8gb 3070 just doesn't have the ram for the job. I really need to try adding some closeup crotch and butt shots to my training but in the past that skewed the results too much that prompting a full body shot was hard, but I only recently realized you can specify how many times it trains specific images and these last few posts reminded me I should try that.
>>35309 Gradient Checkingpointing was the biggest vram saver option I found.
>>35309 If you been doing the base 40 repeat for every training then yup... That's bit excessive. My 4060TI with 16gb does 6-9 batch size at around 15gb of Vram and FP8 training. I have read that if you do all the VRAM optimisations use Cudnn and FP8 you can do SDXL on 8gb no issues. And I believe it because lowest I can run with basic settings in like 9 gb of VRAM. Also keep in mind that you don't need HUGE pictures, the training seems to be better the more sizes you got.
Not a lot of images for an imageboard
Pony is basically cheating at diaper content
The lora definitely doesn't know what I guy in a diaper looks like XD
>>35316 I'm sure we can get that guy who spammed every single thing they made on that (bing?) image generator to spam again.
This diaper has pockets! I think I'm going to retire the dude prompts...
>>35344 This is a common issue with prompts including "standing" and "wearing" in men. I just put "pockets" "hand in pocket" in to negatives and that usually does it. Also "dashion, posing" generally fixes it also.
Anyone have any experience adding AI generated images with bad tapes and trying to teach the concept so that it can be used in a negative prompt?
>>35353 It's easier to teach the correct presence of tapes. SDXL doesn't actually react well to broad negatives, you should only negative prompt specific things. "Girl", "Sunset", "Potato"... etc. The reason the tapes are wonky is because the AI doesn't know what they are supposed to look like correctly. This is why we train the models - we want to add concepts and finetunes we want to add specific informations about a specific concept.
>>35084 Like... 12 versions later I think I'm getting the LoRA where I want it. No longer biased with style. Still not getting the camo pattern with green and blue present reliably nor that "leather tag" which was the design of 2016(?) goodnites for boys. However it did learn this twunk/athletic body also, so I have to add negatives for that.
>>35357 What have you been adjusting over those 12 iterations?
>>35365 Primarily the dataset and it's repeats (and amount of epochs)*. Little bit timestep (currently 50 - 1 000) and then trying to find the best Rank and Dim. This is important because you need to have as high Rank as possible, but not more than is needed (or it learns irrelevant things). *This is difficult because 5 was too few, 6 was weak, 7 was bit too much. However more epochs changes this a bit. Another thing I been doing is that I been trying to spot unnecessary features it picked up. Since I used adult dudes wearing goodnites as the dataset it started to learn the masculine features. Then I cropped, deleted and added compensating pictures trying to balance it out. Oh... And biggest thing I figures out. Variety of resolution is better than high quality high resolutions. I get way better results by having pictures ranging in resolution. Smallest dimension I have is 632 and biggest 1240. It seems like smaller sizes prevent the oversized diaper that doesn't scale down to the subject's size in proportion, and it locking in to a specific area. Like past few iterations I been able to generate without issue so that the diaper can be mostly out of frame or covered by a shirt, and it still holds the correct details and form. Earlier on it was dead set trying to push it in to the frame causing all sorts of proportional issues. However detail bleeding is still an minor issue. Like bedsheets or shirt pulling that camo pattern into itself. But generally this seems to go away with adjusting prompts. (Which I'm worse at than in training). Also I can't figure why it wants to keep putting breast pockets into shirts. The picture posted is from last iteration to the earlier pictures.
I'm close to having something worth sharing. There are a couple issues that I think I am addressing in the current run. If they are sorted I'll post the lora for everyone to use.
(1.35 MB 800x1200 Normal_patties_00130_.png)

>>35377 It's out on civitai. It's never going to be perfect, but given the complete lack of any decent diaper xl lora or checkpoint I figure it's good to get it out there. I've got a laundry list of things I have noticed it is overtrained on, but usually those can be fixed with negatives. Please check it out. If you find any weaknesses, let me know and I'll add them to the list if I don't already have them noted down.
>>35461 Nice, are you gonna upload it to civitai? If you can inpaint a bunch of 'fullbody view, medium distance' generations and post them here, it would be appreciated. Im hoping SD3 lora's will have better quality and detail, but we will see.
>>35521 Nah, I need to generate a ton of training images so I can get a more randomized set of faces before I'd think of uploading it. It looks way too much like my training images at this point..
>>35536 I am by no means an expert, but I find captioning consistent names for redundant training people effectively hides the uncaptioned uniqueness of the particular person.
(1.21 MB 1024x1024 Notbad.png)

(1.12 MB 1024x1024 Trainingsample.png)

Try this lora out, the trigger is "ddiap3r" or "fullbody view of girl wearing a diapar,ddiap3r" So i been using a simple vroid poser to make a dataset, gonna work on making custom diapers lora's for pony by re-texturing it, It's not bad, and the 3D model effect doesn't bleed into it. I'll also try making a Paci Lora this way for pony. Enjoy i cleaned out my mega of old models and threw it in there. https://mega.nz/file/QetT0AxA#wKJfhyvdCGxw4NU-TypycGBa7I0lmA4cRtRHOPICcEQ
>>35584 Yep, I mostly have solved my issue with faces. You really can't tell where my training data came from now. Next will be getting colors to be a bit more random, or making it not so hard to prompt a color.
This thread is falling down the catalog pretty quickly. I don't have any content to contribute but someone should start a new one.
>>35644 Try setting adding training images where the face is partially covered/cropped out, but pacifer visible. I been wortking on a tape gag lora and that helped a lot with preventing it training facial features. Also I think blurring faces (but not the tape) helped. I also threw in few cartoon faces with taped mouth for good measure. After I realised that I started to utilise it in trying other things and it helps a fair bit. I been also trying to make a welding mask LoRA and I realised cropping tactically and having the welding mask in different parts of the picture helped. Give it a try. If it doesn't work for you then just revert back to your old means.
>>35647 How do you handle getting images that are cropped, but still large enough to be useful in a training set? Or do you just grab an image, crop out part of the face, and just leave it at 1024x1024 and just notate in the caption that it's a cropped image?
>>35648 My smallest training image on SDXL has been 380x380 and I had no problems. I make variety of different aspect rations. In photoshop I just crop them (do some adjustments to and quickly remove things like text which I know to mess up things) then scale so that biggest dimension will be 1024. By my experience small traning images do not cause issues AS LONG AS THEY ARE SIMPLE. Because keep in mind you can do random cropping while training when using Kohya. I don't use any captions at all. So I can't tell you anything about that.
>>35644 Kohya has a color augmentation option under advanced. It supposedly shifts the hue a bit during training to weaken the specific color link of the objects. Might be an option. I have used it as I have been focusing on a white diaper. >>35646 Be the change you want to see in the world.
(23.47 MB 2560x7680 old_00001_.png)

(21.95 MB 2560x7680 new_00001_.png)

>>35388 Adding less perfect shots at lower repeats really expanded the flexibility (check out the old vs new for the ballerina set). I'm up to 104 images now. Still a few issues that I have noticed, but I should be able to straighten up. Usually it's either a typo in a caption or one image that really confuses the training. The sock/thighhigh fetish was due to Base SDXL having no idea what a thighhigh is. Thigh high socks, or knee high socks cleared it right up. I spent a lot of time trying the words I was using the captions in prompts to see if Base SDXL understood them already. Ass is out, Buttocks is in. Really helped, but took forever. I haven't figured out the nipples through the shirt issue. There's only a few boobs in the training set and I captioned them all with nipples, breasts, topless. It appears I lost a bit of 4 tape consistency which I think I can get back by adjusting the close up repeats. The 60 new images watered down the close up to other ratio.
>>35046 >>35013 >>34897 Only remotely fappable pics itt. Aspies need to stop spamming 30 variations of the same shit.
I been working on Abu's King diaper. Using Prodigy now which is good, but I struggle to get the fine details reliably. But I been able to keep it versatile.
>>35658 Very good! Can you post somewhere your LORA or your checkpoint file?


Forms
Delete
Report
Quick Reply