growth editionold thread: >>10628103>NAI LEAK SPEEDRUNhttps://rentry.org/nai-speedrun | https://rentry.org/sdg_FAQ | https://rentry.org/sdhypertextbook>FAQ, EMBEDS, HYPERNETWORKS, MODELS, LORAshttps://rentry.org/hdgfaq | https://rentry.org/hdglorarepo | https://gitgud.io/gayshit/makesomefuckingporn>GUIDESPrompting: https://rentry.org/hdgpromptassistInpainting/Outpainting: https://rentry.org/drfar | https://rentry.org/inpainting-guide-SDUpscaling: https://rentry.org/sdupscaleAnimation: https://rentry.org/AnimAnonWildcards: https://rentry.org/NAIwildcardsCreate Embeddings | Hypernetworks: https://rentry.org/simplified-embed-training | https://rentry.org/HNSpeedrunCreate LoRAs: https://rentry.org/lora_train | https://rentry.org/lora-linux-troubleshooting | https://rentry.org/2chAI_LoRA_Dreambooth_guide_english>NEWS AND UPDATES + SD GOLDMINE (RENTRY COLLECTION)https://rentry.org/sdupdates3 | https://rentry.org/sdgoldmine>LOCAL INSTALLNvidia: https://rentry.org/voldy | https://github.com/AbdBarho/stable-diffusion-webui-dockerCPU: https://rentry.org/cputard | https://rentry.org/webui-cpuAMD: Native: https://rentry.org/sd-nativeisekaitoo | Docker: https://rentry.org/sdamd | Onnx: https://rentry.org/ayymd-stable-diffustion-v1_4-guide>PAID BROWSER SITEShttps://novelai.net/>AI BOORUhttps://aibooru.online/>CLOUD HOSTED INSTALLColab: https://rentry.org/sd-colab-automaticPaperspace: https://rentry.org/865dy>RESOURCES4chanX Catbox userscript: https://rentry.org/hdgcbEssentially everything:https://www.sdcompendium.com/Artist styles: https://rentry.org/anime_and_titties | https://rentry.org/artists_sd-v1-4 | https://files.catbox.moe/9wgqx9.jpgBig index: https://pharmapsychotic.com/tools.htmlTags: https://danbooru.donmai.us/wiki_pages/tag_groupsTag effects: https://pastebin.com/GurXf9a4Mix Recipes: https://rentry.org/hdgrecipes>Related>>>/h/hdg>>>/aco/sdg>>>/g/sdg>>>/e/sdg>>>/vt/vtai>>>/vg/aids
>>10634522
was trying for koonago, but got this instead. not really my cup of tea, but here ya go, /d/.
>>10634341https://files.catbox.moe/6ooh1d.pngIt's all the same just rolling.
>>10634528>laura will never pin me against her tits and belly using nothing but the strength and mass of her monstrous throbbing column of a cock>I'll never be able to make a nearby toaster cum in sync with me
>>10634657oh, so its homebrew lora nvm then.
>half a year in>still no headless lora
>>10635038You can complain or be part of the solution.Even a braindead monkey can make a LoRA
Is it feasible to make a LORA for something outlandish like cock transformation? Where the amount of usable pictures is in the low double digits? I gave it the ol' college try, but not having trained any LORAs before, I had no idea what to do, and while there were some similarities to what I wanted, it was all shit.
>>10635052possibly usable for inpainting, but the smaller the dataset the more likely artifacts from the base images will start to crop up, and the harder to get it to go in the first placealso theres a shitload of cock transformation pictures; e621 lists 1.1k; getting non-furry ones might be an issue though
>>10635052>>10635068oh also make sure you fully tag images, the hyperfusion guy ran into that. I'd suggest making variants depending on the stage of CTF. Intact Chest, Fully Absorbed, Face, etc etc, just ballparking.>The key to tagging a 100k dataset is to automate it all. Start with the wd-tagger (or similar booru tagger) to append some common tags on top of the tags scraped with the images from their source site. Then I trained a handful of image classifiers like breast size, ass size, hyper/not hyper..., and let those do the heavy lifting. Finally convert any similar tags into one single tag as described in the tag docs.>For a long time I was plagued with sub par results. I suspected maybe the data was just too low quality, but in the end it just ended up being poorly tagged images. Sites like r-34 tend to have too many tags describing an image like "big breasts, huge breasts, hyper breasts" all on the same image. This is not great for a model where you want to specify specific sizes. Using the classifiers I mentioned above I limited each image to a single size tag for each body part, and the results were night and day.>In order to determine if a new model is better than the last, it's important to have some standard prompts that you can compare with. x/y plot is great for this. Just keep in mind that the seeds between models will be totally different, and you likely need to compare dozens of images at a time and not 1 to 1. It's also important to compare new models against the base model output to make sure what you are training is actually having an overall positive effect compared to the origin model.
>>10635068Non-furry, by half competent artists, and ones where the pictures are easy to split up into sequences.>>10635073So I should make one LORA per stage of transformation, is your idea? I could try that.Is there anything in particular I should try to do with the training script settings? I set it to 512x512 and 1400 iterations or something, since I saw those numbers tossed around. I have no idea what I'm doing.
>>10635082nah I've never made a LoRA, just sharing what I've heardmultipanel is doable but rarehttps://civitai.com/models/52028/breast-expansion-lora-oror-lora
>>10635052I made a mararion lora the other day with just 46 pics I scrounged up from pixiv and it's not overfit and ever works reasonably well. More generic CTF has way more contentDefinitely do make separate datasets for each stage of transformation though, to give yourself more control over what you get and also help the consistency
>>10635089>>10635082Oh and since you're new to this, train at 768 not 512, with default bucketsizes of 320/960. Assuming your gpu can handle that. This will let you gen at slightly higher resolutions and generally lora will give better details than those trained at 512.For iterations, I'd say most concepts learn between 1500 and 3000 steps. For CTF I'd definitely aim closer to 3000(it is how much I did for mararions), just make sure you got at least 50 pics in the dataset to avoid overfitting. I don't fiddle with learning rates or other settings so those numbers are assuming defaults for the lora_easy_training_scripts thing
https://files.catbox.moe/zdcuto.pngI love being able to generate transformation sequences to my own likingbless the onwards march of technology
>>10635046Not meI'm on AMD and it still can't do training
>>10635083>>10635089>>10635098Thanks, I'll keep that in mind and see if I can get something good out of it.
>tfw you will never be as endowed and caked as your creationswhy even live
>>10634870I do recommend that new furry model, I think it's better then the furry kemono mix one at foreskin, not getting as large balls but I'd require more fiddling with it.
>>10635128anon that's the whole bakery jesus christ
>>10635168what can I say I have sweet tooth
>>10634522Inflations? I see expansions are already very good.
>>10635179That's okay anon my weakness is horse dongs.
>didn't manage to download the personality excretion lora before civitai nuked itVery frustrating
>took a break from SD for a couple of months>someone made a ponygirl lora>someone else made a goblin lora>combining the two generates one of my niche fetishes: ugly goblins enslaving cute girls to use them as mountsNeato. Still need to refine it, but I'm impressed with the results so far.
>>10635227Pretty fuckin' good, man
Brain is too small to comprehend the instructions and now I cant make big ball futa images
>>10635389What part are you having trouble with anon?
>>10635389>Download furry model>Mix with non furry to tone down furry>Make large ballsCan't help you for hyper though.
I wonder if anyone remembers that cute redhead from some time ago? I forgot her name
>>10635461Something starting with R
>>10635222tone down the network you're deepfryin her
This is harder than I thought bros... But I think I'm learning
>>10635496>01274what in the bloody hell GPU do you have
>>10635462Ringo!! That was her name
>>10635500I have zero clue what those numbers mean or what I'm doing half the time but it's a 3070
>>10635520it means you've made nearly 1300 images in a day
In general, what's better, more epochs or more repeats? For more 'aesthetic' styles such as starykrows or angstrom, would it be better to use more repeats or epochs?
>>10635224There's one in the gitgud link in the OP; it's just called anal jelly
>>10634620How
>>10635601epochsrepeats only served as a workaround for epoch start overhead in earlier training scripts, and also to 'weight' parts of your dataset as more relevant. But in practice they only serve to make the lora overfit way faster
I have been trying but completely unable to generate my picrel fetish. Any advice. Any success?
>>10635500>>10635570You can quickly rake up numbers if you do a lot of XYZ plots; I'm closing in on 25000 since most of my prompting consists of just doing huge plots and then picking out the best which I then edit in photoshop and then upscale with img2img
>>10633325 Was this made with some kind of "monster manual" lora? Does anyone have it?
Can anybody make one of these with a black schoolgirl growing into a hyper-proportioned Venus? Bonus points if it’s in a school setting
>>10635720Thanks anon!
hypnotized, bound, expanded and bred, bad end?
what affe ts the quality of the gen? the CFG or the sampling steps?
pixai i keeps giving fluctuating quality on the gens, i can't tell if it's the CFG or the sampling steps affecting the quality
>>10635948cfg is how "creative" the AI is allowed to be, the higher the cfg the more strict it is with trying to adhere to propmts. Sampling steps is how much it takes to gen whatever you're prompting, quality wise you'll get dogshit with most samples without being in t he 15-25 steps range, but they all even out around 40-50+ with no real quality increase. You're sampler is detailing your prompt and how quickly it gets made, honestly just use a cfg from 5-8 and sampling steps at 20-25 to be fine with most anything.
im gonna train a model tonight but im not sure which image folder to use.i have done one already of trap boys in body harnesses. the chastity belts are a little strange but you get the idea.
>>10636097the same model can gen from behind also
>>10636099did a similar thing but with 2 having sex. reults were a little messier
>>10636100more than 2 and the results go to absolute AI nightmare land.
>>10636101so what i'm leaning toward next is this style. i don't know what it's called. on pixiv in japanese they all it either "hidden bondage" or "prosthetic bondage." i have 30 something images of this fetish i could train a model on.
>>10636105but i also have something similar like this where it's basically the same thing but on all 4s like a dog. and also girls in bitchsuits. i have plenty of those images i could use also. both photos and hentai.
>>10636109another one could be humans getting milked like cows... i have a whole folder of traps wearing harem girl clothes... so im not sure what model to train today.
moocraft
>>10635518Judging by her art, Amitie is from the same series.Maybe do something like this with herhttps://rule34.xxx/index.php?page=post&s=view&id=7900253Thick, hung and nude
are there any LoRAs for making images of tinies trapped in a giantess' underwear?
>>10635052>>10629882Maybe were'on the near road, Tried nearly 45 (most of theme were shitty manips) images of cockhead and limited success.since it's quite abnormal concept, may need to take very long, hard time training.
>>10636105https://civitai.com/models/47815/exoskeleton-restrained-with-anatomical-issues
>>10635389Here. Most people aren't playing with the sketchy, penciled traditional media style.But I like it.
>>10635390Honestly like where do I even start and download? Not gonna deny my small brain status but theres a bunch of links in the op and I barely know how to use github
>>10636268https://rentry.org/voldyTry that one, homie. That's what I used and it worked flawlessly.
It's never "skipping leg day," when you're carrying around that kind of weight.
>>10635111colabs are free
>>10634522Is there a reason I can't perfectly replicate something I made 10 minutes ago?I copy paste the .txt, same ENSD, same clip skip, model, double check everything and they always look slightly differentI can't replicate it if I try immediately after, but after some time passes it happens again.Anyone know why this is?
>>10636310probably xformers guys, nevermind me
I'm just going to use the explanation "portals" to explain the umage
old image with the first gen app i ever got, too bad the thing went to trash by limiting the amount and going full SFW, hope that doesn't happen to Pixai
>>10636332I'd scream too if i felt an eye forming in me, am i the only one seeing that?
>>10636340those are just the first 32 iterations. I am currently at 100 and it gets a lot worse
Has anyone (can someone) try a monoboob? Uniboob?Character does not have to be a cyclops, there are just very few pics of this.
>>10636214Catbox?
>>10635776My advice is training. There are many guides for training lora.
>>10636200i don't have an account so i can't see it but i guess this is your way of telling me somebody has already done it. i shouldn't be surprised.
>>10636171If anyone struggles with the hat, this sprite sheet shows her without it for her losing animation
>>10634522I've recently gotten started with all this, having gone through Voldy's guide. I've been recommended AOM3 and the Meina mixes as models and I've been liking those results, and I've been dabbling with LoRAs for artstyles and concepts.What's next in terms of learning tools with which to generate images?
>>10636522Probably controlnet, I haven't dabbled in it myself, but it seems exceptionally good at what it does.
>>10636426Do you have any collections of this img?May worth trying w/ 20~30 of 'em
Can anyone help a retard who knows barely anything about computers out? I keep getting this error:OutOfMemoryError: CUDA out of memory. Tried to allocate 21.73 GiB (GPU 0; 8.00 GiB total capacity; 2.33 GiB already allocated; 3.52 GiB free; 2.36 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONFDoes anyone know how to fix it?
>>10636495>hundreds of models and loras uploaded every day>i don't have an accountMake one.
>>10636577Lower the image dimensions or reduce the batch size.
>>10636587I don´t think the problem is the image sizeI used to easily be able to make large images, but now suddenly I cant even make a 256 by 256 one
>>10636590Have you tried restarting your machine?
>>10636600Yeah. sometimes it works, sometimes it doesn´tLike I said I really don´t know much how this stuff works but half an hour ago I couldn´t make images 256 by 256 and now I can do over 1000
>>10636541I do not have that many unfortunately. 1 breasted pics are really uncommon. I think I have a couple more saved but I can't find them.Just these crappy ones.https://files.catbox.moe/8j18h8.pnghttps://files.catbox.moe/pcobpr.jpghttps://files.catbox.moe/6pr6w7.jpghttps://files.catbox.moe/g868fi.png
>>10636306Colabs now kick you out for doing SD after 10 minutes.
>>10636498amazing, please make more
>>10636577>Tried to allocate 21.73 GiB (GPU 0; 8.00 GiB total capacity;Sounds to me, like the image dimensions are to big, or the upscaling-parameters push it to be to big (on my machine HiresFix uses way more RAM, than img2img upscaling).Perhaps also, if you have multiple GPUs, perhaps the program uses the wrong one?
>>10636346Post it
Getting some proper futa sizes can be such a pain sometimes since most of the loras either give too small sizes or are misshapen mushroom abominations.
>a thunderous erection
getting consistent cock sizes is hard
>>10636838Looks pretty soft to me.
>random extra armsNot mad though.
>>10636214Traditional media is a great prompt.
>FWOOMP>GROOAN>GURGLEdon't mind me, just soaking up mana to ferment potions with later
>>10636462https://files.catbox.moe/rlch63.png>>10636860I'm having a ton of fun just making characters in different styles and outfits. I hate what it means for, y'know, long term creative output and corporate abuse and all that, but it's really fucking fun to play with.
>>10636809I would, if I knew how to make GIF of my img2img loopback, but it is too much hassle
>>10636171Pose ref
Bump
>>10636838I lora'd a girl with a dick, the same dick, and my sizes still aren't always consistentAlthough like hands it's a lot easier to notice when you're... focused on it
you'll most likely have your hips crushed, but it'll feel nice while it happens
>>10637102
>>10637308Now add cock.
>>10636565How did you get it to gen two distinct characters? Inpainting?
>>10637580I did use inpainting but not for that.Just character loras for the series and appropriate tags.
>>10635052>>10635068>>10635073>>10635082>>10635083>>10635089>>10635098>>10636184This is the best I could do... There's a resemblance to what I want, but it's very rough.
>>10636916Retard excuse
>>10637669not bad for 512x512try to do some img2img upscaling to add more detail to the frames, it might look much better
Anyone know any loras for /d/orses?
>"Hope you like cake anon! I tried a lot of new recipes!"
>>10637808I made one some time ago, here you go https://files.catbox.moe/0y4y0d.safetensorskeyword is 'dorse' but it barely does anything on its own so make sure to include a bunch of relevant sexual tags
>>10637810Holy fuck that looks amazing god I wish I knew how to use this stuff
>>10637767Do you know how much I should prune the tags? I did two passes, one where I tagged several aspects of the image relating to the ctf itself, and one where I used an auto-tagger, removed the obviously wrong tags, and added ctf. The latter gave me better results, I feel like...
>>10637820Definitely do autotagging. Tagging manually you'll invariably forget shit and under-tagged dataset is no good. After autotagging, other than tossing stuff that's obviously wrong, you can also condense tags(for example, between 'breasts' and 'huge breasts' only keep the huge breasts tag), but that's optional and I didn't bother for most of my lorasPruning/removing tags is odd and I don't fully understand it. I tried it once with satisfactory results but I don't think it was necessaryAs for adding your own custom tokens, don't bother adding more than one. Loras are very odd with new tokens. They range from barely doing something to being completely useless, possibly due to some of them being contaminated/culled during nai training(such as 'paizuri', that token does nothing and is literally untrainable). So just add 'ctf' at the top and keep 1 token in front when shuffling (honestly not sure if this is even needed, I do it just in case)
>>10637833Nice, thanks a lot. I'll give it a shot at 768.
Has anyone dealt with this? I've been trying to get Stable Diffusion running and keep getting this:NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type.
>>10636565is there a lora for those delicious tummies?
>>10637876get a better vae
I remember playing a CYOA "game" which was basically just a kind of checklist. I vaguely recall you were supposed to be entering a kind of competition, where if you made it to the end you'd win a prize corresponding to the limits you've set, and that would include stuff like how you can get fucked, bdsm stuff, but also stuff like body modification etc. I vaguely remember each thing you added supposedly had some effect, though there was no real game there, you just made the choices and it gave you a list of what you chose.Anyway, I can't for the life of me remember what the fuck its name was or really anything helpful about it, and I feel like I've dreamt it up -which I have not-. Does anyone recognise what I'm talking about? Pls send help
>>10638106
>>10638107
>>10638109
Falling Devil too guro for /d/?
>>10637982Makoto Uno lora resulted in that.
>>10637810I saw another different dorse lora on civitai today but idk if it's any good. None of the results on their page looked as good as this.
>>10638106>>10638107>>10638109>>10638111Nice gens. Esthetic trap/futa LoRA?
>>10638283>tgirls
>>10638169>>Makoto Uno lora resulted in that.>look up the gayshit list>the hdg repo>the mega links in the hdg entry>dont find it>remember I had a training set abandoned from the days I still malded trying to get a hypernet/embed off the ground>train a lora with it>it's mediocreYou can't just mention that Lora exists and leave a brother hanging with no link.Here's my attempt btw: https://files.catbox.moe/iv808w.safetensors
I miss the corruption XY sequence guy.
>>10638439https://civitai.com/models/66940/futanari-style-unomakoto
>>10636879>Model: abyssfutacan't find it online, is it your own mix?
>>10638449>couldn't find it either on civitai because its search function sucks monkey fuck as of the past few monthsMuch appreciated.Should have searched for futa, my dataset wasn't 100% futa. I guess I can use both now.
>>10638452It's a mix of multiple models some based madlad spent some time researching mixes to get good cock and ballsAnother madlad went and compiled the recipe for future prosperity, also provided a link to the finished model but without a premium mega account you're seriously better off mixing it yourself.https://mega.nz/folder/fM5lBY5A#4Mk--7W6S2-acfUZ-3PYrgThe archived post: https://desuarchive.org/d/thread/10620880/#q10621205
>>10638477Spoke too soon, I thought you were asking about the kemono mix but nevertheless abyssfuta is part of the components and is there in the archived post.drive.google.com/file/d/1m20HBXPhE7ZcRPmPeQcyrrmD_degZPBV
>>10638468>catbox_0lawzz>0lawzzkekshe comin'
>>10636819>>10636860>>10637479Got catboxex/models? Love me kemonomimi girls with big cocks.
>>10638468Whatever's generating this it's good
>>10638546https://files.catbox.moe/k1tegt.png same prompt for most of mine.
>>10638695nice details
>>10638813Wow.
>>10638813Impressive
>>10638523
>>10638940Now who the hell put an entire raccoon up their ass?
>>10638956That's just a regular tail
>>10638958catbox please?
>>10638813Add_detail lora on civitai is some black magic tier of just adding nice stuff to an image without frying it. Also incredible work.
>>10638996thanks for the tip, i got it
>>10638813kek
>>10639028
>>10639041
>>10639046
>>10639047
>>10639047You use any loras for this one? I'm curious.
>>10639052https://civitai.com/models/52023/erect-horse-penis-lora
>>10638523Catbox?
>>10639041Futa chan has Free food
>>10635108Noice progression.
>>10634620>>10634528>>10636875Man, how do you guys get such good detail and anatomy and size and everything? I've been going for like a month and I thought I was a minority in wanting to make oversized tiddies and dicks. There's like one hyper futa lora on Civitai and I've literally never been able to get good results with it compared to the boobs I get with hyperfusion100k, which gives me decent ones sometimes.This was one of the last images I made on my first day and despite the anatomy being completely fucked on the boobs I ended up really liking it and wished I could get a fixed version of it. Is it all just gacha? I'm using a1111 on an AMD card with some loras but I can never get as much detail or quality as everyone else is getting.I can show some of my best recent gens for an example in some extra posts.also based breast expansion themed OP, never thought I'd see the day I got some representation
>>10634522halp someone, i hired google colab and made the fucking whole process to install unfiltered stable difussion but it doesn't know what is hentai it only shows me amorfic figures what can i do?
>>10639325Some weird black spots on the left boob and I'm still suffering from the blurriness and lack of color that seems to be commonplace for newbies. Using AOM3a3.
>>10639330and I'm guessing some inpainting could fix the hands and face hereI've used inpainting to fix faces just fine before but hands have evaded me. I unno what I'm doing wrong, because oftentimes inpainting just places what I'm specifying there instead of cleaning up what's already there in a way that makes sense to the iamge.
>>10636498 Is there a docking lora?
>>106393253rd guy hereFirst I got a bunch of loras that really emphasize ass size. throwawayjm's hyper loras at civitai and the ponkotsuu lora from gayshit help a bunch with that.Second, rerolling until I get something worth working on.Third, optional, I additionally used my OC homebrew lora which is trained to have a respectable badonkadonk out of the gate.And finally, but most importantly, iterative editing, and a lot of it because huge rising asses from the front are an absolute bitch to make.>Ass not big enough?>Paste into gimp>Clonetool outline of ass an inch or two above its original emplacement>Paint in the gap>Shape ass with warp>(re) apply perspective tricks with the background: paint outlines for the floor and ceiling, hint at correct perspective with transform tools, and so on.>Paste back into SD>Generate at low denoise>Goto 1It's very difficult to get perfect results as txt2chad specially if you're a VRAMlet, and even if you do, they might not be perfect for you.
>>10639330You're missing a VAE. Back to the FAQs.
>>10639503I agree, even the worst sketch on paint can guide sd, take photoshop and enjoy easy composition :D it allows combining multiple images when you have interesting part, using layers too
wanted mummifcation, nemu took it literally
>>10638523https://files.catbox.moe/fy3qx4.jpg
>>10638813lol
>>10639763Praise inpaint \o /
I'm an idiot when it comes to inpainting - how do i go about fixing the hand on this image?
>>10639829Just run it through img2img a few times with the same prompt.
>>10639643As opposed to metaphoric mummification?
>>10634522Anyone want to take a crack at redoing this in ai? I've been futsing around with another image and have gotten mixed results.
>>10639881You were making your own futa lora for these right?Don't know if I missed you saying it, but are you going to share it at some point when it's done, or nah?
is it an issue to share/ask for advice with photorealistic ai generated stuff with /d/ fetishes in these threads? None of the other boards would really fit, and the /trash/ thread throws a hissy fit if you post anything that isnt furry
>>10639909>None of the other boards would really fit>>>/aco/
>>10636660I just used colab to train two models averaging 2500 steps just yesterday.Stop coping with excuses and make your database.
>>10639876i was expecting something like this, like pixai did
can we add https://holara.ai/ to the paid site section? novelai hasn't updated in forever + it has a free version
>>10639907I wasn't before since it kept being scuffed until that furry model got released a few weeks back and now it's consistent and add_details lora really makes it pop. Narrowed down at least from 250 loras down to 23 and realistically I feel done with it, the furry models mixed in with abyss futa and others basically did the other half of the work I couldn't figure out with the training. I wanna just run x/ys to figure out the better or most consistent one at least, last month or so been prepping to move and now the move is happening so was making lora's while I can of characters since I don't know how to use or think I'll bother figuring out paperspace to train them, but I can at least do gens on that website. tldr I'm autistic and yes at some point but expect cherry picking. in your gens.
why do gameplay mechanic stuff like health bars never work as a tag? it's never worked on any generator I've used
>>10639875catbox please?
is there a repository of /d/ loras?
>>10640043Literally look at the OP.
>>10640046where am i supposed to be looking exactly? i see /hdg/ loras, i don't see any list of /d/-specific loras. makesomefuckingporn doesn't seem to have anything /d/-specific either
>>10640047Civit has a bunch too. Just search a little bit. MSFP definitely has /d/ related stuff, too. You didn't look hard enough. Ctrl+f, homie.
>>10639825>>10639843>>10639858Laura gets bigger every time she shows up and I'm all for it.
>>10640121
>>10634522Added a few more LoRA's/LoHa's to the gay-eye mega. https://mega.nz/folder/fM5lBY5A#4Mk--7W6S2-acfUZ-3PYrgStyles added: Starykrow, Debyjull.Updated Maou11 LoHa with 32 additional images, works slightly better now.Will post a few examples of each style.
>>10640224
>>10640225Starykrow
>>10640228
>>10640229And finally, some maou11.
>>10640232Hope yall enjoy them!
>>10640224king shit, gods work, etc
>>10640224its weird to see my loras in there but glad they are appreciated
Spoke to the guy who made this vore sequence with swallowing, apparently he used a LoRa that's no longer available (he deleted his stuff off Civitai).Slim chance, but does anyone have it saved and is capable of uploading it somewhere? Been looking for this for ages but can't find it.
>>10640356I didn't want to remove them because that'd go against my whole post from a whole back of me trying to archive shit so it'd be easier for others to find, so I just shoved them into the "others" folder.Now I just gotta wait for gayshit anon to put my link in the repo.
Trying to make Mineru but not having much success, anyone have any advice for training furry lora/locon?
>>10640537go check out the /trash/ diffusion thread. they specialize in furry stuff
>try to make a nugget with a caretaker robot>it makes the bot quite challenged as wellThat's what I get for over-relying on loras.
>>10639991https://files.catbox.moe/mbwnyo.png
I know this is the AI imagen thread, but you degenerates seem like the best people to ask: How the fuck do you get ChatGPT to stop appending stupid bullshit to the end of my weirdly specific smut stories?I've been pretty good about steering around the lewds for transformation and weight gain themed stories so it doesn't start freaking out and saying I'm breaking the rules. I still occasionally get some butthurt output about how it can't output stories that are mean-spirited or that promote unhealthy habits, but I usually just regen a few times or change a couple words on my prompt and it's back to generating elaborate stories about anime girls getting fat. The problem is that it usually ends every output with a couple paragraphs of obnoxious bullshit that goes like>and even though protagonist changed a lot, she learned a lot about herself and grew as a person thanks to this personal self-growth and exploration experience. Little did she know, but she had many more fun adventures ahead of her!And it does that shit on nearly every fucking output. Anyone else having this problem or am I just a retard?
>>10640714You can't make it stop hallucinating after writing most of the stuff you asked and there is more space left.With a bit of editing, you can make it write your story bit by bit or ask for it in one go and then ask questions like "describe the thought processes, from the protagonist's perspective, about how he felt when his friend took control of his brain and started to fondle his now huge female ass". This away, you can make it fill most of the stuff it left behind unwritten.
>>10640714Use an unaligned model locally. Llamas won't moralize.
>>10640784Got a guide for that?
>>10640548>>10640551Thank you've cummed so I've turned into a little shriveled husk of a raisin.
>>10640801are you okay anon
>>10640819Poor guy cummed so hard he damaged the language center of his brain.
Anyone knows if there's a list of the artists in Fluffyrock's dataset?
>>10640819Imagine the smell (of her apartment)
the orc/goblin/pigman lora from civit is gettin stale? any good monster loras or animal loras?need some like horse monsters banging chicks, aliens, bugs, demons, minotaurs, resident evil style lab super zombies, etc. wouldn't mind better versions of the orc/goblin/pigman ones either if someone knows of some
>>10635089>mararionno clue what that is, but catbox?
>>10640477https://mega.nz/file/gvomhSQZ#I8PlYb9krLKqjSLhT_F4BLnGG1uy-LQwHbUoihA6wIQ
>>10641050that's a good hole, but this is abit better
>>10641050>>10641075
so there's a lora for mieruko-chan. never seen it myself, but the example images got me wondering if it can output some nice unaware groping scenarios. it can
>>10640914fuckin based, thank you so much anon
>>10641096I know it’s vore, but I can’t unsee leg tongue
>>10640790>>>/g/aicg/ >>10640714>>>/g/lmg/
I like how extremely squishy she looks
>>10639881>>10639965Absolutely sick dicks. Probably the best AI dicks I've seen.
>>10640714If you make it "write in the style of an erotic roleplay chat," it won't do that as often. Only other way I've found is to give it very clear directions about exactly what to write.
is there a futa POV LoRA where the breasts are more prominent? This one likes to shy away a bit when animating. This has other problems too but I'll figure it out.
Nemu's generator is pretty neat. Is there a free AI that turns sketches into images? I'm new to this
>>10641284I don't think there are any free ones. Nemu's is great and simple to use though.
>>10639503 hereI am bored of rerolling "ass, huge ass", trying to get front views of a hyper-caked up character, only to get her to do a 180 and moonwalk her tush into my face.Pasting back and forth into gimp to try and get something coherent got old, so I decided to bake a lora for that.Did you know, almost nobody draws asses from this angle large enough that you can see them rising up past the sternum? Building a small set not composed of 99% furry took way longer that I had patience for. The activation word is "frontass" (as a mirror/pun on "backboob"), it also hijacks "ass" and "huge ass" to force a forward facing angle.https://mega.nz/folder/Q3pVGabY#mlSw1oDvs2vgrHyNc9h15ABy the way, I decided to organize the loras I've made, in case somebody actually wants them.https://mega.nz/folder/UmB1VYrb#1B4wsw1EQ3Rw6121BKDUZg
>>10635179please more
>>10641133Looks like soft caramel. I want to chew on her.
>>10641133NOOOO I DON'T WANNA BE CARAMELActually, being caramel isn't so bad
>>10641133Is there a lora for onaholification?
>>10641133Need more 2hus like this, pls do more
>>10641435I can't seem to make this look better in a timely manner but here you are
>>10641486Multiple, actually. There's sex toy transformation on civitai and a couple onahole models you can find in archives of /ddg/.
>>10641506can i see the ones you didn't deem worth posting too? ngh
>>10641432gods work, will download
>>10641509well, the thing is, there isn't, its just dozens of i2i iterations of that one except with random penises or weird striations. This is the best cherrypick out of ~20 and then passed through i2i about 3 cherrypicked times without editing
>>10641519now if only ai could make me into that
>>10641432>hyperlips>my biggest fetishany good examples?
>>10641470lawd have mercy, more of this
>>10641545
>>10641548Not bad, danke. Guess it doesn't get much bigger than that?
>>10641550They do get slightly bigger than that, you basically use "hyperlips" as trigger.Not at the computer anymore but here's my original post: https://desuarchive.org/d/thread/10628103/#q10632373
>>10641553oh shit, that was you? I enjoyed those images greatly, certainly gonna take a crack at that.
dongle
holy shit I love AI so fucking much
>>10641486I think I got it here, it's called "sexToyTransformation_v1"
>>10641133Got a catbox for this one?>>10641504nta, here's one of mine from a few months ago
>>10641704it's the combination of loras that really gives the squuishiness
>>10641711Oh the slimegirls LoRA, that's a nice combo. ty!
Ah, here is where this belongs. This honestly went surprisingly well for how many loras I jammed in here, with surprisingly little inpainting necessary.
Futa on futa doggystyle Minimal inpainting needed on this, nice!
this thread needs more expansion sequences, I wanna see torn clothes.
Still not perfect, but for a raw txt2img I'm much happier with the quality of each gen now.>>10639521Thanks! I got a VAE installed and my generations look a lot better, much more color and life. Can't believe I've been generating this long without one.>>10639503Ohh I see, looks like I'll have to go into learning to mix in some image editors to set something up. Thank you so much for the detailed response. I'm guessing to get the balls to that size in your image you did inpainting and editing stuff, too? I have the hyperballs lora from gayshit but can't get anywhere near that size. Again, thanks man.
>>10641982and with the magic of inpainting I've gotten some of my favorite gens yet, doing a set of gens slightly based on the Miku prompt from the BEReaction Lora on Civitai.>>10641900I was trying so hard to get Remilia on the ground with hyper breasts bulging out of her torn dress, but no matter how hard I inpainted the pink on her dress just seemed to blend in with the tits and the AI didn't understand. Not really a sequence, but that being said I'd really like to make more at some point.
>>10641982>I have the hyperballs lora from gayshit but can't get anywhere near that size. Again, thanks man.I happen to also have trained that one, which I also used. Beyond what I already mentioned, for testicles specifically I find that mentioning "testicles" before "penis" in the prompt helps with avoiding multiple penises. I presume this happens because of inconsistent tagging in most models where the testicles aren't mentioned unless they're prominent, so most models conflate "penis" with the whole package. So the AI may full package on "penis", then get confused on "testicles" and draw it as a mutation elsewhere.Other than that it's rerolling and editing what catches your eye. Here's a catbox of a cherrypicked old txt2img using said lora while I was testing it.
>>10641998What's the highest ass you've made so far? I'm into that.
>>10642026Purposefully? Probably >>10641432, because the others kind of cheat with the angle. I rolled this once while aiming for something else but it looked pretty bad to me at the time. Maybe I'll rework it later.
>>10642026>suddenly, out of fucking nowhere, the most blessed idea crosses your mindyeah, scratch that.
Working on SWB Loras recently.
>A futa's final orgasm before she's turned completely into an onahole
More cock fucking.
>>10642071noice
>>10642055>>10642037>me, an internet sicko:Hahaha...yes, hahahaWell done anon, love a girl taking up so much space.Although I really love the ones with a huge dingdong, saved all the ones you've posted before. Have you had any luck with making them sitting? Does the ai know how to do that?
nami got hit by an attack, with unexpected results
>>10642071amogus
>>10642083perfection, I would love more
>>10642132How about some 1st person view?
>>10642140incredible
>>10642107>Although I really love the ones with a huge dingdong, saved all the ones you've posted before.The dark-skinned one? He's supposed to be a feminine boy, but I'll take that as a success lmao. I guess I lean more into the "draw a girl without breasts, call it a boy" school of thought.>Have you had any luck with making them sitting? Does the ai know how to do that?It appears this lora I made can't do that. There were a few pics in the dataset of humongous seated asses taged as sitting but i guess it wasn't enough, and it wasn't my primary target anyway. This took heavy editing and the hyper bottom heavy lora.
>>10642199Great, I went and uploaded the previous upscale, not that it made much of a difference.
>>10642199Actually I figured it was a boy and was cranking it as such, but you never know what people will reee about. Honestly that's an ideal body, amplifying a big dick swagger with a swaying ass. >>10642201looks like sitting to melove it, will always appreciate it when I see it.
new anon hereWhat in your opinion is the best small penis lora, and the best chastity cage lora?Tried some but got mixed results
Original:https://files.catbox.moe/v7tyez.png
>>10642280jadf is that you?
>>10642300Hi
>>10642322Didn't expect to see you here, welcome to this lil weird corner of /d/. Have you been having fun with AI?
>>10642353You can find my AI stuffs here:https://www.dropbox.com/sh/c5uetaox9f4qify/AACKk__s_tD0_wMJPEG6PHM5a?dl=0
>>10642358Very nice stuff! Do you plan on eventually uploading it to your twitter or somewhere else?
>>10642306I love her...
Anyone know of a LLM that can do vore and similar properly?
>>10642358can you try generating hasumi from blue archive, like that one comic
whats the best model for traps/otokonoko? i was trying waifu diffusion but it just wont give me a bulge reliably
>>10642210>Honestly that's an ideal body, amplifying a big dick swagger with a swaying ass.I love hyper stuff but more importantly conceptualizing how hyper bodies could possibly function, and the hyper ass with hyper balls connection one of my favorites.For starters, the weight of the ass alone serves as a counterweight to the humongous testicles. So the person isn't just anchored to the ground and can walk, albeit precariously. The extremely wide hips make it a challenge, but they're needed for there to be enough room for the scrotum to hang confortably, the fist-thick spermatic cords that feed a seminal vesicle larger than their head, and enough bloodflow and musculature to keep it going.Speaking of musculature, there should be a lot hiding down there not just on the legs and ass, but also around the belly to help it contract during orgasm, squeezing that vesicle and using the vacuum to pull that thick dense baby batter up into the vesicle from those heavy cum tanks, ready for the next spurt.They say hyper is monke tier but for me it's the thinking man's fetish. Although I gloss over a little over the nutritional requirements of such a semen factory and the square cube law, but you can't always be on top of everything.
>>10642872all I can say is, you're not wrongYou're not wrong at all. And that's another thing I like about your guy, the way his overdeveloped abdominals belly out over his dong.
>>10642877>And that's another thing I like about your guy, the way his overdeveloped abdominals belly out over his dong.It is a trait that kind of popped up as I was developing him, which I initially disliked but came to rationalize it as being positively,completely packed with of cum, and needing that kind of abdominal development not just to squeeze his innards hard enough to blast away a pound per contraction, but to keep himself from doing so until the time is right. And I guess a bit of all that ass fat makes its way to the front.God I wish that was me.
>>10642947Biggest tum yet!>God I wish that was me.Same brother, same. Thanks for sharing.
Not my picture and no clue about the prompt sorry.Does anybody know/have a Model or Lora that can such "bestiary pages" ?
>>10642071
futas should have human sized cocks
>>10643055>consistent veins that somewhat make sense>peeled back foreskinnice cock
coffee time
>>10643017Make her a futa and I will cum buckets.
>>10643226Reminds me of a lake I went to a couple times as a teen.
>>10643191
someone pick an image from this thread so i can slap the logo on it
>>10643254>>10641133
>>10643243
new thread: >>10643313new thread: >>10643313new thread: >>10643313
