Tiny speedo editionPrevious thread - >>3090593 (OP) #*Request Thread: >>>/y/airt>To make a new thread, wait for page 8 at bump (and/or image) limit in catalog.>ONLINE GENERATORShttps://tensor.art (NO LONGER SUPPORTS NSFW)https://tensorhub.art (NSFW version of tensor, It's no longer free sadly)https://civitai.com (NO LONGER SUPPORTS NSFW, use the link below for that)https://civitai.red/ (Civitai new NSFW page)https://refineforever.com/ (Free but limited options and probably slow)https://tungsten.run/ (10 generations free every day, but the premium it's cheap)https://novelai.net/ (PAID ONLY)https://pixai.art/es (MAYBE?)>GOOGLE COLAB FROM LAST THREADhttps://colab.research.google.com/github/gutris1/segsmaker/blob/main/notebook/Segsmaker_COLAB.ipynb>METADATA & TAGS EXTRACTION/LISThttps://www.metadata2go.comhttps://rentry.org/airt-rs3https://rentry.org/povlist>/airt/'s RESOURCEShttps://rentry.org/airt-rs1 (FAQ, etc.)https://rentry.org/airt-rs2 (Setup)>QDIFFUSION (LOW-END PC)https://github.com/arenasys/qDiffusion>LOCAL FORGEUIhttps://github.com/lllyasviel/stable-diffusion-webui-forge>LOCAL REFORGE *recommended*GitHubhttps://github.comPanchovix/stable-diffusion-webui-reForge>LOCAL MODELShttps://civitai.com/models/184813/animeconfettitunexlhttps://rentry.org/5exa3 (hll+fluff) *OLD*https://gitgud.io/gayshit/makesomefuckingporn (use wayback machine) *OLD*Naixl:https://civitai.com/models/833294/noobai-xl-nai-xl>RESOURCESTraining: https://rentry.org/59xed3 | https://github.com/derrian-distro/LoRA_Easy_Training_ScriptsTags: https://danbooru.donmai.us/wiki_pages/tag_groups | https://danbooru.donmai.us/related_tagAnimation: https://rentry.org/AnimAnonWildcards: https://rentry.org/NAIwildcards4chanX Catbox/NAI prompt userscript: https://rentry.org/hdgcbNAI API/prompting in auto1111: https://github.com/Metachs/sdwebui-nai-api>RELATEDFurry (Male) AI art: >>>/trash/msdgAi art request thread >>>/airt/
So Homosimile Anima has been released: https://civitai.red/models/2708104/homosimile-anima?modelVersionId=3041842
It is still very bara focused, but has an OK style out of the box.
Basic prompting that feels OK:Positive: masterpiece, best quality, very aesthetic, extremely detailed, score_7, Negative: censored, bar censor, worst quality, low quality, lowres, score_1, score_2, score_3, jpeg artifacts, bad anatomy, watermark, artist nameThe suggestions on the web page for Homosimile are pretty decent in so far as sampling, sets, etc.. are concerned.I do feel that CFG is best left high, I have been using 6.
To be honest I haven't been all that impressed with Anima so far.You are capable of somewhat more complex scenes and small English writing. It also has more characters trained in, even male ones, which is a nice step.
But even with complex written sentences I have struggle with some character blending. I am also getting issues with sex scenes and basic poses that you think would be trained in. I cannot figure out daisy chaining oral sex, for example. Maybe I just don't know the prompting well enough. But then what is the point of having a better text encoder if it wasn't encoded for sex acts?Will keep at it and report here and there when I figure out something worth sharing.
love this series at the moment
>>3092015 :thumbs up:
naughty bunny boys get punished>>3091988> I haven't been all that impressed with Anima so far.same>>3091989>what is the point of having a better text encoder if it wasn't encoded for sex acts?honestly. if the dataset wasn't so opaque we'd be miles ahead already
1, 2, 3 edgelords edging on the edge of a tree
>>3092020To be fair it is pretty early in development for Anima. It took a year before Pony came out for SDXL, and awhile for the fine tunes and Loras to get good. Another year or so for NoobAI (illustrious), and then the same amount of time for it's fine tunes to catch up.Will just have to keep watching
Tried to see if I could push NovelAI to do Pixar-style 3D. Was not particularly successful lol, but this one turned out okay.
>>3092035>Linguingimy neck just snapped, and i just chuckled
>>3092025So why do you like gaping so much? Why are they always bara characters?
>>30920381. clean assholes are always nice, so imagine if you could see it very blatantly.2. women are vanilla and anything short of muscular gets blown away by the slightest gust of wind.
>>3092038>>3092039Lol, this interaction.Sometimes I venture into gaping a bit, but I don't like prolapse. Same thing with bara. I mostly gen twinks/twunks but sometimes I'll mess around with some big pecs and thick thighs.
Been on a bit of a Naruto thirst lately
>>3092040I was honestly curious, and English may be their second language. They have a CivitAI account I believe, and it is literally all just different bara characters gaping. They have been posting them for a long time, like since Pony got popular.At first, I honestly thought it was a troll that didn't like AI SLOP, so they were trying to poison any AI thread. Then I realized they were a strange person with a strong fetish.
>>3092043Yeah, I believe you. I didn't know they had a civit account. They've been posting in these threads for a long time now haha.
What is green buzz? why are three different kind of buzz on civitai right now?
>>3092020
>>3092052bunny boys are such sluts... anyway that's it for these two for now
Happy Father's Day!I bring you the controversial daddy from Final Fantasy X, Jecht
>>3092051CivitAI has been pretty open about their issues taking card payments from traditional companies. Those companies have been spooked by the ability to produce shota/loli, amongst other illegal or immoral pictures and videos they don't want to be linked with. I would note that the payment companies are doing this to many businesses, not just CivitAI.CivitAI had to break the website into two pieces, CivitAI and CivitAI Red. On CivitAI you purchase green buzz with regular payment cards to generate highly restricted images. On CivitAI.red, you purchase yellow buzz with crypto to generate x-rated images subject to the basic censorship of 18+/consensual/no beastiality.On either website you can earn blue buzz by interacting (upvoting/censoring/ads) to generate restricted images. I will leave the larger argument about censorship via large profit driven multinational corporations to other boards.
>>3092054The "I want to put a baby in you!" pose is always hot!
I did some art of an older Davis and Lou from Digimon (based on their appearance from 02: The Beginning) both wearing loincloths from their country.
>>3092063
>>3092064
Liking the posting, great to see people having fun
>>3092063I have no imagination for clever scenes
I do like making pretty boys though
>>3092080>>3092081Sometimes I do scenes, but a lot of the time I just want simple hot guys lol
Playing around with your trademark cosmic lora :)
Saw someone talking about Anima for couples on the last thread, thought I’d give it a shot
>>3092085https://civitai.red/models/599757/velvets-mythic-fantasy-styles-or-flux-pony-illustrious-zit-anima?modelVersionId=1982022Very good Style as well. On the dark side, and the backgrounds suck. But I love the lines it makes for male muscles
>>3092143https://civitai.red/models/966483/96yottea-style-or-illustrious?modelVersionId=1082928I like this one for more abstract style. Good with men ( a lot of style loras are so built around females that it feminizes everyone). The uploaded picture is dark because I like it that way, but can do bright pastels as well.
>>3092144https://civitai.red/models/1398477/hassaku-or-shiiros-styles?modelVersionId=1580764This is one I found recently. Need to add the "realistic" tag for full effect. Also, you may want it at 0.8 instead of 1. 1 can be pretty strong and pastel colored, and a little too much female. Makes great femboys though.
>>3092145https://civitai.red/models/1324670/sinozick-or-shiiros-styles-or-niji?modelVersionId=1495543I love this one as well. Very dark with great lines like >>3092143"Smoothlines", but it isn't quite as grainy and better backgrounds and lighter colors. It is more female based though, which is odd because it does bodyhair great. But easy to get girl eyes and facial structure even with a honking big cock.Need to add "gradient, spot color" for full effect.
One last style: https://civitai.red/models/599757/velvets-mythic-fantasy-styles-or-flux-pony-illustrious-zit-anima?modelVersionId=1373674"Mythic Portrait" is an old one, but for whatever reason it plays well with character Lora. It does a nice job when one of my other favorites won't do. Worth having around.
>>3092143>>3092144>>3092145>>3092146>>3092147These are all awesome! I'll play around with them this weekend, thanks for the recommendations.
>>3092185if you end up starting an account somewhere, drop your username, will happily follow
>>3092194Haha, thanks. I've been considering it, maybe I will. I'll post it here if I do. :)
any success on making a looping gif?
>>3092224I have never revisited video gen once I realized that wan 2.2 didn't work with images created in illustrious.
>>3092229I mean, it works. But you can't use the style Loras, which I am in love with
Also, I only have 32 gigs ram and 16 vram, which is on the low side for video, so I don't experiment with it
Slow here again, figured I would share something
Jog peoples creativity maybe...
>>3092224Try this in loop mode
>>3092237I can't put the url for some reason
>>3092233Good taste in Pokeboys.
>>3092240Turo just being everyone's whore compilation
>>3092241
>>3092211magnificent
>>3092302>>3092303I really have to work on a goblin twink boy one of these days...
>>3092307Lol, I have a bunch of twink/twunk ones but I'm always nervous because I'm not sure what /y/ mods consider too young or not.>>3092309Still learning comfy but stuff like this makes we wanna figure out the hi-res adjustments.
Anyone got any tips for making goblins in Novel AI? I've tried before but they always come out a little too anime looking and I want to get something more along the lines of BG3, more classically "ugly" looking
>>3092316I haven't used NovelAI, I do all local gens. I assume Novel doesn't have loras? Because that would be the easiest way. I tend to lean towards anime anyways, but hopefully someone else has some answers for you.
>>3092316Try strengthening the goblin tag.Here was the character prompt for picrel:boy, 3::goblin::, green skin, bald, fangs, foreskin, large testiclesSome other tags you might be interested in are warts, sharp teeth, claws, skinny, belly, wrinkled skin, earrings, pauldrons, loincloth, fur trim, etc.>>3092321NovelAI unfortunately doesn't and never will support LoRAs. There are a few reasons, but if I'm not mistaken, one of the main ones is that since NAI gens are completely private and unrestricted, LoRAs would allow people to create a lot of dicey things like photorealistic CP.It allows you to upload reference images for characters and/or styles and is trained to recognize a very wide range of concepts though, so LoRAs aren't that necessary anyway.
>>3092322And here are some clasically ugly goblins raping Ike for funsies
>>3092312Thank you for the goblins lol. That particular picture (>>3092309) was a result of this Lora: https://civitai.red/models/646411/f2-midjourney-v61-meets-flux-illustrious-sdxl?modelVersionId=1385235used on this checkpoint:https://civitai.red/models/140272/hassaku-xl-illustrious?modelVersionId=1617798and not a result of any hires upscaling. Just FYI
>>3092303you best believe i'm getting a handful of that
>>3092316have you try using the character reference option?
>>3092321Yeah NAI doesn't use loras, I hear about them a lot in the anime generals I frequent though>>3092322>>3092323Wow thanks! These all came out great, and even seem pretty modifiable too for if I wanted skinnier and lankier or chubbier gobs with the bellies>>3092327I have not but this is a good idea, never tried the referencing tool because I can never tell when I should use that or the vibe transfer for what I'm looking for
>>3092335The bodies are coming along well but faces are still tough, they tend to come out super ugly, like way more than I want and borderline orcish with how big they are. Hard to get them to a point where they're in that ugly-cute range where they're still somewhat handsome in their own gobliny way. But doing the face slightly out of frame helps, will mess with other tags to see if I can make em a bit more attractive while still preserving that goblin face look.
Liking the different goblin stuff
I always felt people miss just how creative you can get with image generation
It takes a lot of free time to practice though
but it can still do some amazing things
Kinda related, I'm vibe coding a mock dating app with forge neo implementation and Agnes API where the characters can send you photos, here is this dude sending me photos from a naked beach
>>3092403Damn, wish more AI text apps were like this.
>>3092403I was hopelessly embarrassed the one time I tried to do a roleplay chat with a bot before. Literally could not handle it. But I wish you the best.
Just uploading a few pictures to keep the thread fresh.
hope everyone has a good weekend
and can stay cool
>>3092429Lol, that's so funny. Part of the reason I post less is because I've been using SillyTavern too much, it's a lot of fun once you get the hang of it.
>>3092430>>3092431>>3092432Hope your weekend is refreshing and fun as well. <3
I don't come to /y/ very often but may I ask what are some good twink or crossdressing loras for Wai Illustrious?Thanks my bros.
>>3092437Unless you have a specific character in mind that is not trained in (you can check here: https://animadex.net/ ; any character under 125ish entries will be poorly trained) you should NOT be using Loras.It is a very beginners trap to think you need a lora for everything, when really prompting is key. For a twink character, try these prompts:Positive: "teenager ; slim body ; 19 years old" / add "toned male ; pectorals/ biceps/abs" to give them some muscles, if you don't want soft.Negative: "facial hair ; mature ; old ; fat/heavy/hefty ; (might need: female/women/girl/makeup) if it pushes too feminine for you."crossdressing ; crossdressing_(mtf)" should work just fine as well.Then actually describe what they are wearing. There are a ton of female clothing prompts, feel free to experiment. Some that should work well:"thighhighs ; bare midriff ; lingerie ; male underwear ; see-through "clothes" (see-through shirt/top/bottom/pants) ; choker ; lipstick ; eyeliner ; makeup ; dress ; straps ; o-ring ; bodysuit ; etc. etc. etc. Really, the matrix should be really good at making crossdressing young men out of the box. Don't forget you need to describe your character in some detail, hair style/color/length, eye color, skin tone. You should not be using a lora for anything like that.If the character keeps coming out too feminine for you, you can add some male only positive prompts like "body hair". Or add some female traits to negative like "makeup ; high heels ; pussy ; breasts ". That should help keep the matrix on the right path. Don't forget the basic : "1boy ; male focus" as well in positive.I did some cross dressing here if you want to take a look: https://civitai.red/posts/22561932
>>3092437>>3092442To piggyback off my friend here, "bishounen" is a great tag for twink/twunk. In general, you can go on danbooru and check images that interest you and look at their relevant tags.For example, "crossdressing (mtf)" on danbooru, check the posts for the images, check the wiki for the explanation and related tags: Genderswap (MtF), Crossdressing (FtM), Trap, Male maid, Male playboy bunny. Obviously you don't care about the crossdressing ftm tag, but the others could prove useful.Have fun and share what you generate with us! <3
>>3092450this is literally my boyfriend except he has a mullet and brown eyes
NovelAI devs just posted a teaser for the upcoming v5 (catboxing because it's a SFW 2girls image):https://files.catbox.moe/htonuc.png"An interior and background sample from the in training v5 model. Keep in mind model will still train for a lot longer. No inpainting used, pure text to image with natural language + tags mix."NAI's biggest flaw/local's biggest advantage over it for illustrated art (besides price) for a while now has been its lack of understanding of backgrounds/objects, so this is a big bridge-gapper. They've also been showing teasers for edit model functionality (the "describe the changes you want to see" feature others have had for a while now) and its improved ability to do things like text and comics.It's been lagging behind in some pretty major ways despite being king in others, so I'm extremely excited for it to catch up and become more well-rounded.Ability to do backgrounds could also be huge for some games I'm workshopping (jury's out on whether it will be good enough for standalone backgrounds, but it should at least work for sex scene backgrounds where the details aren't the focus).
Love Naruto, teen or older <3
>>3092462>>3092463I want to peel his foreskin with my tongue and then try to choke myself on his schlong as cup his heavy balls.
>>3092462>>3092463Naruto is a 10/10 iconic design. Sexy and cute when he's a shota, teen, or DILF. Crazy about all 3. I subbed to NAI just to gen self-indulgent smut with him, but of course it's good for messing around with other characters and things too.
>>3092466>>3092472Lol, I'm glad you agree. Naruto is super hot. All the shonen protags are to be honest. Naruto, Deku, Denji, Yuji, they can all get it. :3
>>3092473Shonen protag is almost like an inherently attractive archetype, but Naruto just feels like the peak of that design philosophy.
>>3092476Bratty, horny and ready <3
Little bump with some old shit. Hope yall are having a good weekend. <3
>>3092500hope you are too mate!pic by AlbriechArwe
some good posting, but went quiet again
Hope the heat wave didn't damage any gpus
dick
note
Your periodic NovelAI shill update:v5 will be able to generate images with transparency, which will be nice QoL for making things like visual novel talk sprites:https://files.catbox.moe/29pttw.png
>>3092547The era of gay vn slop is upon us!
>>3092559You know it brother. I still plan to make my original VNs without AI, but I'm planning some porn parodies of games/anime with large casts, so that would be way too many characters to draw (not to mention CGs, those would be a nightmare).People seem a lot more willing to accept AI in porn games than regular games, and most of the ones out there even have the generic slop style.
>>3092560>People seem a lot more willing to accept AI in porn games than regular games, and most of the ones out there even have the generic slop styleThat's because porn, for better or worse, has lower standards than "normal" stuff, and people don't mind that while getting their rocks off.
>>3092563this was a test using this tool https://www.youtube.com/watch?v=1GBgacrYT9o i'm not used to comfy but the tool it's really easy to use and fun, and compatible with anima too
From a while back when I had a thing for college Andy lol
>>3092547I love it
>>3092600This is awesome, were you able to prompt the 4 way split like that or just combine 4 pictures?
>>3092624Andy is cute. Not sure about the gymbro abs, but still very attractive
Posting some Hiro Hamada. No incest though
The idea of making a giant robot for sex is hot, especially if it looks so benign
A secret sex robot to sneak past your parents
Testing Krea 2
>>3092635
>>3092636
>>3092638hot, more please
gpt 2 is fun to crack
The Krea stuff is interesting, will have to give it a shot
I don't think that Krea is explicit out of the box though?
but maybe it can come up with some interesting nsfw stuff
I do like being teased
>>3092683It's not trained in a lot of sexual poses or even flaccid penis but it has surprisingly good cocks even in the base model (this is not, but I don't remember the name of the model), nudity is censor by a method people already patched.
>>3092684>tight legginsHOT, would love a front view with a bit of bulge
>>3092536how do you prompt for these sort of rainbow/holographic clothing?
>>3092697anima has no idea what holographic is but you can say just that: prismatic boxer briefsyou can say rainbow too but you risk rainbow bleeding into creating an actual rainbow somewhere
Your periodic NovelAI shill update:Lots of interesting teasers for the work-in-progress v5 posted lately, but the most interesting were the ones showing its abilities to generate complex scenes and multiple characters.Here's one:https://files.catbox.moe/544sm4.pngThis is without any inpainting/editing/upscaling/reference images/etc., just pure txt2img. (It does use NAI's character positioning feature though, which lets you assign characters to a spot on a grid.)To show how much cherrypicking is involved, the dev also posted the batch the image was picked from:https://files.catbox.moe/qr5wh8.png
>>3092715Here's another teaser from today showing its ability to generate 8 distinct characters performing different actions with complex text placement:https://files.catbox.moe/ovre07.pngAs before, pure txt2img except the character positioning feature.And another batch to show how consistent it is:https://files.catbox.moe/hicl6c.pngThe dev who made these teasers posted the prompts for both. They're both girthy but simple--a bunch of Danbooru tags (mostly to describe the characters' features since they're NAI OCs) + one short prose sentence.v5 also has an increased token limit, and at least 8 character boxes. Orgies incoming?
>>3092694He is an oldie but goodie. Was playing around trying to copy an old picture from Gelbooru i liked (which I can't find now). Original picture was from behind, so I didn't really make any from the front.
>>3092697So illustrious can do rainbow clothing, just put "rainbow" in front, ex: "rainbow boxer briefs". It does have a high chance of bleeding into other parts of the picture though.
>>3092697Not exactly show what you mean by holographic, but this was made with "holographic boxer briefs". (in illustrious based model).
how's everyone liking krea 2?
Asking based on the previous conversation around NovelAI up above…What do you all think is the best way to generate images of more than one character and have them not blend together when trying to make NSFW stuff. Is it NovelAI? Right now I only have access to TensorHub and using LORAs is tough for multiple characters. I’ve tried using anima to write it out but it seems it doesn’t recognize many different characters/fandoms. Trying to write it out for WAI Illustrious and Illustrious models also is extremely hit/miss in my experience. Should I try Pony? I’m not sure how to use that one.In all honesty I’m still trying to learn by trial and error with the rest of them too. I’m not 100% sure of what I’m doing most of the time…
Anyone been able to find LORAs for either Mizuki or Lifeweaver from Overwatch? Mostly looking for Mizuki.
>>3092710>>3092718>>3092719apparently the word i was looking for is "iridescent". thank you all
>>3092718>>3092719These look awesome! I'll def be playing around with both.
>>3092724Iridescent is a little off in this style lol. I'll have to try it with some adjustments, it seems promising.
>>3092722It's NovelAI. It gives you up to 6 separate character boxes (although 4.5 struggles with more than 3 characters), each with their own prompt and undesired content fields, and also the option to position them within the image, so it's very easy to avoid bleed (there's also very little bleed TO avoid).The main problem with bleeding I run into is certain elements that aren't part of a character's design, like their state of (un)dress. If I prompt one character to be nude it likes to make the other one nude too, for example. It's generally pretty easy to fix though, you just need to write more descriptive character prompts and/or use more numerical emphasis. You can also just temporarily toggle off one character's box while you inpaint the other.The only time it's hard to avoid bleed is when you're using the character reference feature (sort of like a LoRA, except it's a single image you upload). That feature's currently only intended to support a single character, so it can turn everyone in the image into the referenced character. There are ways around that, but they can be difficult.Once v5 releases, it should be even easier than it already is to avoid bleed. My main worry with v5 is that since the model is going to have a lot more general knowledge, it might have a harder time recognizing rare characters. 4.5 only requires about a dozen high-quality images of a character on Danbooru to recreate them, but if v5 requires more, then the reliance on character reference would increase, which as above has issues with multiple characters... Of course, fingers crossed it'll go the opposite direction where it requires less images to recreate characters (like with v3 --> v4), but either way we'll have to wait for the model to finish training to find out.
>>3092727I’ve read that NovelAI doesn’t allow NSFW content to be generated and has filters to detect it. How are you getting around it/get it to make NSFW images?
>>3092728Don't know where you read that but thankfully it's completely wrong, NovelAI is one of the most NSFW-friendly generators in existence. There are no filters--your generations are completely unrestricted and private (even the devs can't see what you're making, not that they care). You can make guro, lolisho, or literally anything else that the model knows how to make.NovelAI's entire MO is respecting people's privacy and creative freedom, and they don't run ads or have investors (all of their money comes directly from users' subscriptions and credit purchases) for that very reason, so that there's no one pressuring them to clean things up.You might be thinking of the Curated version? Each model gets a Curated version that can't do NSFW, but even then, it's not filtering anything, it just literally doesn't have NSFW in its training data. It's basically a half-trained, non-horny checkpoint.