File: ad9e346e-010a-452f-a7f1-7(...).png (1.71 MB, 1024x1536)

1.71 MB PNG

**THREADLY NOTE**
Need to find a character's template/LORA? https://civitai.com/ | Built-in tags/artist styles (Pony): https://docs.google.com/spreadsheets/d/1m2W-pZEvHuEpfHcNHrxCSr-Aw1mgtUUYho6sz9LChEA/
Need inspirations and/or couldn't find the right words to describe what you want? Try this: https://danbooru.donmai.us/wiki_pages/tag_groups
>>>/y/airt is also good for all sorts of questions/advices regarding AI (and requests).


Naixl resources:

Styles and characters:
https://www.downloadmost.com/NoobAI-XL/danbooru-artist/

Usage recommendations:
https://civitai.com/articles/9695/f-best-negative-prompt-for-noobai-xl

Models from last thread:
https://civitai.com/models/1091817?modelVersionId=1226261

https://civitai.com/models/964011/homosimile-xl-pony-illustrious-noobai

https://civitai.com/models/1313975?modelVersionId=1483194

File: f2b9f609-a831-4b6e-817a-e(...).png (1.78 MB, 1024x1536)

1.78 MB PNG

>>3080872

Anyone have more Zidane/Link/Aang?

Those were goooood.

File: 00041-3288372976.png (2.86 MB, 1296x1728)

2.86 MB PNG

>>3080884
I made some with Kuja not that long ago

File: 00037-436306717.png (2.62 MB, 1728x1296)

2.62 MB PNG

>>3080884
Had issues with the tails, but otherwise came out ok

File: 00001-692738970.png (1.78 MB, 1296x1728)

1.78 MB PNG

SDXL: Diving Illustrious Flat Anime v2.0

File: 00068-3607203195.png (2.03 MB, 1728x1296)

2.03 MB PNG

SDXL HomoSimile XL v4.0

Worked on some hairy twink boys

>>3080892
I do like arrogant cuties. Very nice.

>>3080890
>>3080891
These are amazing. Ffffff~

File: out1.mp4 (3.36 MB, 592x880)

3.36 MB MP4

bit too interpolated but meh I'm just getting the hang of this shit

>>3080919
That's really good, which tool dis you used?

>>3080927
Wan2.1GP with a missionary lora.

Anywhere people share nsfw ai stories? Have had some success and wanted to post it/see other people's.

>>3080893
>>3080894
I love this trend of hairy cute boys so much

File: Mauga.png (571 KB, 1536x1536)

571 KB PNG

>>3081052
I don't know anywhere dedicated to NSFW AI stories (I think I heard of the NovelAI community having a Discord for it or something), but you could try the AI-Generated Text tag on Ao3:

https://archiveofourown.org/tags/AI-Generated%20Text/works

Granted the works in there aren't all necessarily going to be entirely or even mostly AI, since the tag is a bit broad. I have seen at least one account on Ao3 dedicated to posting AI stories though, and surprisingly there wasn't hate in their comments sections (or maybe there was but they were moderating/deleting it), just some people pointing out when the AI did silly things (like going back and forth on a character's pronouns).

File: 00080-1340822936.png (2.09 MB, 1728x1296)

2.09 MB PNG

>>3081053
<3

File: 00081-1340822937.png (2.32 MB, 1728x1296)

2.32 MB PNG

>>3081057
I love letting the matrix come up with interesting poses on its own. So many interesting pictures can be generated if you let some of the guide rails stay low.

File: 00117-2227895788.png (2.57 MB, 1296x1728)

2.57 MB PNG

>>3081058

File: 00152-1935902351.png (2.49 MB, 1296x1728)

2.49 MB PNG

>>3081069

File: 00019-4068850928.png (2.04 MB, 1728x1296)

2.04 MB PNG

>>3081070

Is it possible to run this on a laptop? I have 16GB RAM but otherwise just a normal intel laptop GPU/CPU and no money to buy a graphics card for thousands.

File: 00040-2654263292.png (2.37 MB, 1536x1536)

2.37 MB PNG

>>3081093
For the original Stable Diffusion 1.5 (which is bad at porn) you need around 6 gigs of Video Ram (from graphics card) and 10 gigs regular Ram.

SDXL models (Pony ; Illustrious) which are still the most popular for anime porn, like what is posted here, at least 12 Gigs Vram and 16 regular ram.

The newer models (Flux ; Wan 2.2 ; Chroma) like the ones people are making the 5 sec videos with, would need 24 gigs Vram, and a solid 64 gigs of regular ram.

Keep in mind you also need a solid state drive with a good 200 gigs of space free, or more, depending on how many individual models and Loras etc. you want to have downloaded.

One caveat, the numbers I have posted here are somewhat debatable. But they area a good rule of thumb for quick and error free generation. People can make the matrix function with less, but it adds a lot of time. That time will make it frustrating to use and much harder to learn. As the newer models are distilled down, the hardware requirements will drop some, but it will take time.

There are other ways of using the matrix without owning a $2000 USD Graphics card.

https://vast.ai/ and other sites like it will allow you to rent the GPU processers for a few dollars a day. You will still need the solid state drive space, and a good internet hook up. You will also need to get the matrix programs up and running on your own PC. https://www.reddit.com/r/StableDiffusion/wiki/index/ can help.

You can sign up for a site like https://civitai.com. They will let you generate all kinds of anime porn, and you can have access to the library of Loras etc. that they allow. But, freedom is limited to what checkpoints they host, and what keyword prompts pass their XXX filter. But it can be a good start to see if you actually enjoy it.

Good luck!

>>3081093
even if you could it wouldn't be worth doing, you're better off searching for a google colab that is compatible with the free tier

I spent 2 seconds search and came up with this. I didn't try it so can't promise anything: https://github.com/cagliostrolab/forge-colab

is there a discord bot that allows you to generate gay porn images?

>>3081107
I'm not the one you respond, but Civitai is a miss & hit. At times I use it because it easily generate bad faces with deformed eyes and mouth, more if they try to mimic a more detailed art style. And I know you can fix it with Face Fix or Hi-Res Fix, but they barely fix it.

Sadly Tensor banned nsfw and get models there is a nightmare if you're new. But as one that is old and have several models in bookmark, still is one of the best for generate stuff. It did indeed was able to generate dark stuff at times, all without the bad eyes & mouth that Civitai causes.

File: 893173165692317558.png (865 KB, 768x1152)

865 KB PNG

So I decided to make some AI art of Lyman Banner from Yu-Gi-Oh GX based on that one scene from Episode 46 (spoilers).

File: 893173165692317559.png (961 KB, 768x1152)

961 KB PNG

>>3081130

File: 20c30f39-a7ea-462f-bde2-f(...).png (1.58 MB, 1536x1024)

1.58 MB PNG

Trying to recreate baitbus with chroma

>>3081158
Pretty nice, though the eyes need work.

File: 00005-1564430931.png (3.14 MB, 2048x1536)

3.14 MB PNG

>>3081162
I forgot to use Adetailer on it, damn. Chroma generation time is such a pain on it's own. But i'ts really fun to use

The bottom dude has also six fingers now that i'm looking. .

File: 00005-1244288239.png (2.52 MB, 1296x1728)

2.52 MB PNG

SD

File: 00101-2038577282.png (2.64 MB, 1296x1728)

2.64 MB PNG

HomosimileXLv4.0

File: 00175-1020460874.png (2.89 MB, 1536x1536)

2.89 MB PNG

ReForge U.I.

File: 00131-2632700093.png (2.98 MB, 1296x1728)

2.98 MB PNG

SDXL

File: 7NPTETZHP0CN563F2ZJG71XH20.jpg (167 KB, 832x1216)

167 KB JPG

relatively new to this, but pretty happy and horned up by how these turned out

HomoSimile XL

File: BRDB2PV0F1HSG7CF3MP4KTY740.jpg (189 KB, 1216x832)

189 KB JPG

>>3081223
WAI-illustrious-SDXL

I keep getting duplicate bodies even though I explicitly state in positive and negative prompts for `light-skinned male` and negative `same body, same skin color`. interesting how the wai-illustrious version makes him twinkier, though I can't complain since tauro is insanely breedable to me any body type he is

>>3081125
>Sadly Tensor banned nsfw and get models there is a nightmare if you're new. But as one that is old and have several models in bookmark, still is one of the best for generate stuff. It did indeed was able to generate dark stuff at times, all without the bad eyes & mouth that Civitai causes.
fuck! i missed out this...guess im fine with sfw gens for now

File: Tauro Anal.png (540 KB, 1856x1280)

540 KB PNG

>>3081224
Man, I don't miss living in a world without NovelAI's character boxes...

File: brock.png (874 KB, 768x1280)

874 KB PNG

>>3081298

File: flash.png (1011 KB, 768x1280)

1011 KB PNG

>>3081299

>>3080893
>>3080894
Holy shit these are great. He's so hot.

>>3081343
That fucking face is the best thing.

File: tmprknk9x7x.png (1.82 MB, 1280x1280)

1.82 MB PNG

>>3081355
lol, it's my go to for a lot of my gens.

>>3081416
Prompt, please? He cute

>>3081417
it's a zelolee lora

File: 2f896fc5-cc53-41a5-8fd1-0(...).png (1.05 MB, 768x1152)

1.05 MB PNG

>>3081366
Love the style! How did you achieve that? Looks like a mix of beitemian and szadek loras

File: tmpygrfiupk.png (1.59 MB, 1280x1280)

1.59 MB PNG

>>3081452

Close! beitemian and idorrad, no loras. Danbooru tags work if the artists have enough work on the site.

HomoSimile XL v4 checkpoint

And then just put this in your positive field:
(beitemian:0.9), (idoraad:0.9),

File: 4835010b-6b08-4f58-a644-5(...).png (1.7 MB, 1024x1536)

1.7 MB PNG

>>3081458
I've tried beitemian before, but not idorrad, thanks

File: tmp6_jmp8fx.png (1.59 MB, 1280x1280)

1.59 MB PNG

>>3081475

Gotchu pookie <3

File: 28cd3c93-a51e-428c-bd29-1(...).png (1.78 MB, 1024x1536)

1.78 MB PNG

>>3081498

File: 43cf17c3-d590-40e5-bf3c-d(...).png (1.78 MB, 1536x1024)

1.78 MB PNG

>>3081499

File: 00069-homosimileXLPony_v4(...).png (1.96 MB, 1280x1280)

1.96 MB PNG

>>3081458
How did you get the dick to look so good? (you're not painstakingly inpainting them, are you?) I often get very red pointy tips (from idoraad, I guess). I'd love to see the prompt

This one is from trying to replicate the look of your pics, and the dick is still a bit too scraggly and reddish

File: tmp2ohpxsjh.png (1.86 MB, 1280x1280)

1.86 MB PNG

>>3081503

Fuck nooooo, I hate inpainting and would rather keep generating new ones lol. But all my dicks look fine, occasionally the tips can get red but not often.

Here's my full settings and prompts for these ones:

Euler A, 1280x1280, 30 steps, cfg 4

positive:
1boy, male focus, muscular, adult man, angel, angel wings, short blonde hair, topless, halo, nude, flaccid penis, testicles, (lots of cum, cum string, copious amount of cum, cum on face, cum in mouth, cum on chest, cum on legs, dripping cum, drooling,) abs, obliques, parted lips, asleep, eyes closed, tired, blush, blushing, BREAK laying on clouds, heaven, clouds, lots of clouds, night, stars, galaxy, nebula, space, BREAK cinematic, dynamic lighting, dynamic pose, film grain, depth of field, bloom, very awa, masterpiece, best quality, absurdres, newest, perfect quality, detailed background, intricate details, (beitemian:0.9), (idoraad:0.9),

negative:
lowres, worst quality, bad quality, bad anatomy, signature, username, logo, watermark, artist name, old, bad hands, weibo watermark, patreon username, pubic hair, body hair,

File: tmpq0jewyjj.png (1.8 MB, 1280x1280)

1.8 MB PNG

>>3081505
>>3081503

Might be the Euler A? Softens images somewhat.

The checkpoint can use DPM++ 2M too, but it makes finer lights and more detailed textures. I like the softness of Euler A.

Same settings for a variant, with different positive prompt:

1boy, male focus, muscular, adult man, angel, angel wings, short blonde hair, yellow eyes, topless, halo, nude, penis, erection, testicles, leaking precum, lots of precum, precum drip, precum string, (disembodied hands, handjob, hand grabbing penis, hand grabbing testicles, hand grabbing pec, finger in another's mouth, tongue out, drooling,) looking at viewer, abs, obliques, parted lips, half-closed eyes, tired, blush, blushing, BREAK heaven, clouds, lots of clouds, night, stars, galaxy, nebula, space, BREAK cinematic, dynamic lighting, dynamic pose, film grain, depth of field, bloom, very awa, masterpiece, best quality, absurdres, newest, perfect quality, detailed background, intricate details, (beitemian:0.9), (idoraad:0.9),

File: tmpht_wi2iz.png (1.94 MB, 1280x1280)

1.94 MB PNG

>>3081506
>>3081505
>>3081503

This is a fun prompt for like partially shown cock/underwear stretching:

1boy, male focus, muscular, adult man, angel, angel wings, blonde hair, yellow eyes, topless, halo, white thong, detailed bulge, (penis peek, penis barely visible, lowrise, lowrise underwear, underwear falling off, underwear pulled down, underwear stretch, hand pulling underwear,) looking at viewer, abs, obliques, parted lips, half-closed eyes, tired, blush, blushing, BREAK heaven, clouds, lots of clouds, night, stars, galaxy, nebula, space, BREAK cinematic, dynamic lighting, dynamic pose, film grain, depth of field, bloom, very awa, masterpiece, best quality, absurdres, newest, perfect quality, detailed background, intricate details, (beitemian:0.9), (idoraad:0.9),

Pretty much use the same prompt for everything and then alter it as needed, brackets are my explanations:

1boy, male focus, [age, body type-muscular/skinny or both], [character looks- hair, eye color, body part descriptors, earrings, hat, etc.], [clothing or lack there of- shirt/tank, pants/shorts, topless, underwear, etc.] [if nude, penis/testicles/ass/anus descriptors here and any positioning] looking at viewer, [if topless, i put abs, obliques, here] [facial expressions- parted lips, half-closed eyes, tired, blush, blushing, etc.] BREAK [location and background descriptors] BREAK [I never touch these, they're always in my prompts- cinematic, dynamic lighting, dynamic pose, film grain, depth of field, bloom, very awa, masterpiece, best quality, absurdres, newest, perfect quality, detailed background, intricate details, (beitemian:0.9), (idoraad:0.9),]

File: tmp470ne21e.png (1.61 MB, 1280x1280)

1.61 MB PNG

>>3081503

Btw, this looks great, please share any other gens you do/like <3

>>3081419
Moar of Milo! Childish face with a bara body is my kink !

File: tmpfs7gmsbm.png (2.04 MB, 1280x1280)

2.04 MB PNG

>>3081511

catbox is pausing my uplaods rn, don't know if it's them or me, but I'll try again later. Made a small zip of some Milo stuff for you.

File: tmp64b395fe.png (1.87 MB, 1280x1280)

1.87 MB PNG

>>3081511
>>3081512

File: tmph4ojw19e.png (1.76 MB, 1280x1280)

1.76 MB PNG

>>3081511
>>3081512
>>3081513

File: tmp36qmzqpb.png (1.49 MB, 1280x1280)

1.49 MB PNG

>>3081510
>>3081509

These are hot, very nice!

File: tmpaef_k97_.png (1.94 MB, 1280x1280)

1.94 MB PNG

>>3081511

https://files.catbox.moe/71yk9x.zip

Small collection of Milo pics

>>3081545
Luv you

File: tmpv8mfrblw.png (1.77 MB, 1280x1280)

1.77 MB PNG

>>3081567

<3

File: 00214-homosimileXLPony_v4(...).png (1.23 MB, 848x1272)

1.23 MB PNG

>>3081508
Thanks a lot for the help, by the way!

I also wonder if there's any tips for getting deep penetration to work. Seems oddly difficult to prompt — this one is with "balls deep, deep penetration" (and also I put "just the tip" in the negatives for fun)

File: tmp6yo7kqny.png (1.65 MB, 1280x1280)

1.65 MB PNG

>>3081584

No problem! Always happy to help. I tried ((balls deep, deep penetration,)) and was able to get it pretty deep I think.

File: tmp5l9y23xf.png (1.58 MB, 1280x1280)

1.58 MB PNG

>>3081596
>>3081584

File: 00175-1116263100.png (2.76 MB, 1536x1536)

2.76 MB PNG

>>3081584
One thing to keep in mind, if you are adding penis based prompts such as "erection, veiny penis, huge penis" it will generate with less penetration in order to show you the cocks. Try just prompting for testicles and "deep penetration" instead. The matrix will still make the cocks, but they will be inside better.

File: 00050-1950260554.png (1.8 MB, 1296x1728)

1.8 MB PNG

SDXL

File: 00026-3401500808.png (2.44 MB, 1296x1728)

2.44 MB PNG

Homosimile 4.0

File: 00185-2699777277.png (2.84 MB, 1536x1536)

2.84 MB PNG

Stable Diffusion

File: tmplain1j7t.png (1.36 MB, 1280x1280)

1.36 MB PNG

>>3081602
>>3081601
>>3081600

Love your gens always, can I ask what the base is? I know they're different checkpoints, but they all have the same style and it looks great.

>>3081599
That is true, for mine my only penis based prompt was "erection". So that makes sense.

File: 424cfcd9-7bad-4002-9e01-0(...).png (1.66 MB, 1024x1536)

1.66 MB PNG

>>3081603
That's such a hot concept

File: 00205-1505849530.png (2.86 MB, 1728x1296)

2.86 MB PNG

>>3081605
<3

I usually only use 2 checkpoints for anime.

HassakuXL Illustrious V 1.2: https://civitai.com/models/140272?modelVersionId=1200105.

For Hassaku, the style you are seeing is created with the Lora "Velvet's Mystic Fantasy Style" at 0.8 strength. https://civitai.com/models/599757?modelVersionId=1373674

Note: higher number versions of Hassaku have been released, but I don't like 2.0 and haven't yet tried 3.0

The other checkpoint is HomoSimile XL V 4.0: https://civitai.com/models/964011

The style is created with the Lora "Cosmic Fantasy Art illustriousXL" @ 1.0 strength: https://civitai.com/models/1420520

Honestly, HomoSimile is pretty darn good without any styling Lora, but I do like the darker look and sort of gritty realism that Cosmic Fantasy adds. However, Cosmic Fantasy is best at making actual fantasy portraits. It will actively choke the generation if you are trying to make everyday scenes, and not some sort of knight or cosmic angel.

Note: I do not like Homosimile V 5.0, and I honestly might like 3.0 better than 4.0.

Anyway, I post some of my stuff here: https://civitai.com/user/gtgiygus
CivitAI has all of my prompts and checkpoint/Loras visible for your own use if you want to experiment. Have fun!

File: tmpjvwhbplb.png (1.42 MB, 1280x1280)

1.42 MB PNG

>>3081607

Thank youuuuuuuuu! I love the cleanliness of your images, different from my style, but I'll def gen some in that way. I haven't touched Hassaku since pony, so that might be my mistake.

Cosmic Fantasy Art illustriousXL

That looks super interesting, must be what I'm thinking of, thank you for the info.

Homo 5 sucks, I hate it, not sure what they did exactly, but it is super restrictive and doesn't make gens the way I want without heavy lora influence and even then they are a bit iffy.

I'll check out your civit link, I might finally make an account of my own/share my images, I just prefer to raw save the data in txt documents but it would be nice to have a "portfolio" of sorts, just to share the series I make.

ps. your hairy twinks are hot af :P

File: 00210-1149794478.png (2.28 MB, 1296x1728)

2.28 MB PNG

>>3081651

<3<3<3

There is such a thing as overtraining a checkpoint.

At the end of the day, all of the checkpoints we are talking about are based off of a few "base" models created by cooperate/research entities with large amounts of money to spend and a great deal of computer power.

In our case, the Illustrious model, started with LAION-5B (5.85 billion images, according to google) and added something like 8 million more images from Danbooru.

Small scale operations like HomoSimile don't have anywhere near the resources to make a new model. So what they are doing is, in a way, sort of training a Lora INTO the base model to give it a specific style with a small number of images. For example, bettering male anatomy, which is overlooked for the more popular tits and vaginas.

The (relatively) small amount of images (due to time and money) will then start to outweigh the base, which had the flexibility due to the much larger training set. This means it will start to only output images looking like the smaller set, constricting the whole thing. There is a lot of art involved in the science of training, and the amount of time it takes makes it hard to pivot if something goes wrong.

You will noticed that as the popular, but small scale, checkpoints keep trying to improve, they often become less nimble due to the erasure of the 8 million images they started with to focus on the small set they think they want to aim toward.

What happened to all the male celeb loras literally cant find anything anywhere

>>3081779
got sued.

File: tmp68qty053.png (1.66 MB, 1280x1280)

1.66 MB PNG

>>3081780
>>3081779

Yeah, they're hard to find and even then they're pretty shitty. I tried commissioning a few and they always turned out like ass.

I tried a few times to make loras, following tutorials really closely and they never worked. I'll try again some day, but for realistic models I think it's just difficult without lots of knowledge.

>>3081685
That's a lot of helpful info, thanks! Totally makes sense. I'm pretty happy with my go to checkpoints, primarily Homosimile 4, but I occasionally use others.

File: fa4ea3bb-e4f4-4b85-a366-5(...).png (1.63 MB, 1536x1024)

1.63 MB PNG

Trying to do multiple panels on chroma but the prompt bleeds a bit somehow

File: 8e6dbf86-d5db-4a68-bbd0-a(...).png (1.64 MB, 1536x1024)

1.64 MB PNG

>>3081798

Two things I wanna ask or have been meaning to ask for a while:

1. Without resorting to the yaoi-specific checkpoints (I'm using one of the more "popular" Illustrious merges in the Nova series for now), is there any foolproof way to specify which character should top/bottom? Been doing a few scenes recently involving two generic unnamed characters where one of them is darker-skinned, and the model insists on almost always making the dark-skinned guy the top regardless of the keywords I try. Rarely, it does decide to randomly switch the two around, which is good, but I want to get this more consistently.

2. Is watersports/pee stuff fine to post and discuss in this thread? Just wanna check in advance before I ask for any specific advice on that one too.

File: 00019-3775361829.png (2.74 MB, 1296x1728)

2.74 MB PNG

>>3081940

1) Even using "yaoi specific" checkpoints it can be very difficult. All porno trained checkpoints are still 95% created with straight content, and therefor the "masculine" character will be on top. You can try brute forcing it with multiple gens with low samping steps and no hires.fix or ADetailer and cherry pick from there.

The newer (much more resource heavy) checkpoints like Chroma and Wan 2.2 are better at following specific prompts.

There are also addons for your U.I that create "Regions" to specify which charter will be in which section, which can aid your ability to generate certain poses. Forge/A1111 popular one is "Regional Prompter", though I am sure Comfy has one as well.

2) I have seen a bit of golden shower content here and there, don't think it is against any rules.

Anyone know if there's a danbooru tag for this artist:

https://www.pixiv.net/en/users/12199930

Can't figure it out, can't find a lora either because of the name.

File: bd0367e8-5963-45f6-9eb2-1(...).png (1.85 MB, 1536x1024)

1.85 MB PNG

More chroma testing

File: 00085-1199169010.png (2.28 MB, 1728x1296)

2.28 MB PNG

>>3081984

Head to danbooru and search "unknown". Check the artists that pop up with "unknown" as part of the name. If you find one that matches the person you are looking for and they have more than 150 hits, copy the name as part of the prompt and add "style" at the end. For example: "beitemian style" without the quotes, and give it a shot.

If they have less than 150 hits, there was probably not enough training data for it to matter and it won't work.

>>3081988

Oh I looked, none of the "unknown" artists are that art, it just surprises me none of their art is on danbooru or it's under a different name, but I can't find any other names they use.

They have tonsss of art out there, so it seems weird it's not on danbooru.

File: 00021-1585532453.png (2.73 MB, 1728x1296)

2.73 MB PNG

>>3082007
If you a name for them in the future (or anyone other artists that interest you) check it against Danbooru using the same rules. The more hits, the better it will work.

File: 00161-1937810722.png (2.27 MB, 1296x1728)

2.27 MB PNG

>>3082008
<3

>>3081945
nice anymore like this

File: tmp5235wlnl.png (1.74 MB, 1280x1280)

1.74 MB PNG

>>3082017
>>3082008

File: 00043-55372638.png (2.8 MB, 1728x1296)

2.8 MB PNG

>>3082021
An experiment I wasn't satisfied with. I wanted darker skin and lighter hair. But the matrix kept making the skin with tan lines, or straight out two toned.

File: 00009-1669316352.png (2.85 MB, 1728x1296)

2.85 MB PNG

>>3082033

>>3082007
there are a lot of gay artists that aren't on danbooru or are very underrepresented

if you want them to be baked into future models you gotta upload them yourself

>>3082035
as an addendum: if you do, please don't upload low quality stuff, and avoid watermarks if you can

garbage in, garbage out

I noticed some artists trained on danbooru stuff got screwed up badly on novelai lol

>>3081607
Holy shit, thats hot!!!

>>3082034
nice they look great what ai did you use and what prompts
i would love more

File: 00010-596536064.png (2.8 MB, 1728x1296)

2.8 MB PNG

>>3082039

Positive: very awa, masterpiece, best quality, absurdres, newest, perfect quality, detailed background, intricate details, 1boy, male focus, nsfw, 18 years old, teenager, brown skin, blond hair, dyed bangs, asymmetrical bangs, green eyes, wearing elaborate tribal style clothing, serious, toned male, nipple slip, testicle peek, sandals, anklet, decorative belt, detailed designs and multiple straps, patterns of rhombuses, accents of tan orange brown and black, grasslands, sunny, blue sky, clouds, looking at viewer, <lora:cosmic_fantasy_art_illustriousXL:1>

Negative prompt: facial hair, mature, tanlines, lowres, worst quality, bad quality, bad anatomy, signature, username, logo, watermark, artist name, old, bad hands

Steps: 50, Sampler: DDPM, Schedule type: SGM Uniform, CFG scale: 7, Seed: 596536064, Size: 1152x864, Model hash: f44814ee79,

Model: homosimileXLPony_v40NAIXLEPSV11, Denoising strength: 0.4,

lora:cosmic_fantasy_art_illustriousXL: 4b8fab5d02,

Adetailer: Anzhc Face seg 640 v2 y8n.pt: d473e8bccc, ADetailer confidence: 0.6, ADetailer dilate erode: 4, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer use separate sampler: True, ADetailer sampler: Euler a, ADetailer scheduler: Karras, ADetailer version: 25.3.0,

Hires CFG Scale: 7, Hires upscale: 1.5, Hires steps: 35, Hires upscaler: R-ESRGAN 4x+ Anime6B

>>3082043

Will respond in HuCow thread

>>3082047
thank you for sharing your prompt.
Im trying to understand how to make things and i got comfyUI to run,

but im still finding it difficult to get anything working.
Do you by chance have any recommendations? I cant even get similar results with identical prompts i found on Civitai.

And isolated prompts never have good proportions... I want to get into making AI art, but i cant seem to get a basic humanoid that isnt horribly deformed (and your result looks so nice). Im overwhelmed and dont know where to start/why it isnt>>3082047
working.

File: 00124-2180683740.png (2.56 MB, 1536x1536)

2.56 MB PNG

>>3082068
So there is no easy answer. I have never learned how to use ComfyUI, so I can't help you there. I started with reforge/forge and never looked back. Forge is always a couple of steps behind the newest stuff, but my system can't handle the newest stuff anyway. Honestly, I would try a different UI, like forge, and go from there.

This link is what got me started: https://www.reddit.com/r/StableDiffusion/wiki/guide/

I use this wrap around program to make things easier, would highly recommend it: https://lykos.ai/

Just start copying people and you will learn eventually. https://civitai.com/ is where I browse. Lots of good ideas.

Good luck and have fun!

>>3082069
I really appreciate it.

I know erotic art is niche as is, gay erotica even more so. Finding out the how and why, when you are savvy enough, is a real deterrent

anything to help is appreciated; its hard to troubleshoot when you're walking blind
-
So thank you. It really means a lot
I'll give that link a go

Cheers!

File: 00163-525228130.png (2.06 MB, 1296x1728)

2.06 MB PNG

>>3082070
As long as your system has the resources you need to generate pictures in a timely fashion, you will be able to learn. If your system is slow, it will take so much more time it will be very hard to practice.

Download Stability Matrix from the Lykos.AI site, It will help simplify things. It can even browse CivitAI for you and download Loras, checkpoints, and more. Stability will help you install different U.I. next to each other. You can have Comfy and Forge up and running on your system in no time.

After that, it is practice practice practice. Keep googling things you don't understand. Ask questions here.

The two noob traps I see people fall into are 1) Using far too many Loras and 2) Filling prompt boxes with useless stuff. Try just generating clean without any loras to start with. The newer checkpoints I talked about, like Hassaku and Homosimile, work just fine without any loras at all.

>>3082047
nice they are hot i would love more

File: A3D_03835_.png (1.23 MB, 896x1152)

1.23 MB PNG

>>3081945
Yeah, I figured that Regional Prompting would be the way to go at some point, but the problem is that I'm not sure how that would fare for the composition and posing I've got at the moment (one guy directly in front of the other guy he's riding), when I always thought the two characters were meant to be as separated as possible for that to work. I did figure out a solution eventually, but it did take a fairly cluttered prompt in the end.

Picrel is an example from a few days ago of what I mean regarding the composition. I did a cum version since I still wasn't sure if posting WS stuff straight away would be a good idea, but I have some Catbox links to demo that as well.

https://files.catbox.moe/zix9t0.png
https://files.catbox.moe/u8e0zq.png
https://files.catbox.moe/ff8a80.png
https://files.catbox.moe/6dl81s.png

And for the part 2 of that post, I keep wondering if there's any way in general to control things like flowing fluids and where they should be landing or flowing towards. In my case, generating a lot of WS, I always get very inconsistent results in terms of piss streams. They rarely, if ever, tend to land where I'd want them to land, and sometimes they don't even land back down at all, either going off-screen or just abruptly stopping midway through. I don't know if I wanna resort to inpainting since I haven't had the best results with that so far, and it's also extra annoying that I only started out with AI gen altogether some time after CivitAI already banned watersports content completely. What do?

File: 00054-2729169121.png (2.55 MB, 1296x1728)

2.55 MB PNG

>>3082108
I honestly think the piss came out really well. The only thing I can think of would be to prompt where a puddle would appear, but I suspect SDXL would not get it right often. Maybe next gen Chroma/Flux would understand enough.

There are regional prompters that work with user created shaded areas, instead of just the squares you can make in basic regional prompting. I have never used them, but I think they work OK. Will take a lot of practice though.

>>3082120
I see. Again, I've done nothing regional yet, so I'd have to start from the ground up there, but that sounds promising.

And thanks, by the way. Male watersports tends to turn out well more often seeing as you obviously have dicks to aim with, but again, my problem is more that it's not consistent sometimes where the streams themselves go. Let's say I wanted it to land on the ground or a certain part of the body, for example. I could try all the keywords i wanted, but a lot of the time the model still won't know what to do with that and either send the piss all the way out of view or just abruptly stop it mid-frame. It's even more frustrating to work with for female watersports, but that's obviously out of scope for this board.

FWIW, I'm using an Illustrious-based checkpoint (Nova Anime3D XL there) under ComfyUI. I haven't tried anything to do with Chroma or Flux yet, so I'd be curious where to start for that.

(And lastly as a bonus: https://files.catbox.moe/qq7tfd.png)

File: 00025-41233399.png (2.35 MB, 1536x1536)

2.35 MB PNG

>>3082121

>my problem is more that it's not consistent sometimes where the streams themselves go

You won't get anything better without more work. "Not always consistent" is basically image gen's subheading. Cherry picking mass produced gens is a legitimate approach, just skip any hi-res and ADetailing steps until later. Forge makes it easy, I am sure Comfy has a way to activate those detail steps after the fact as as well.

>I haven't tried anything to do with Chroma or Flux yet, so I'd be curious where to start for that

https://huggingface.co/black-forest-labs/FLUX.1-dev

The good news is Flux (and Chroma, which is really just a retrained version of Flux) operates pretty much the same as SDXL (which Illustrious is a retrain of). Flux doesn't have a negative prompt (Chroma does have negative, at the cost of computation time), and requires you to make a detailed paragraph in English about what you want to see. This paragraph should included poses, styles, colors, shadows, etc. etc. as well as the basic characters and actions SDXL requires as well.

To put it more basically: SDXL needs you to be short, concise, and on topic. Flux needs you to be almost verbose, but you still need to avoid "purple prose".

The bad news is Flux basically doubles your system requirements. 24 gigs Vram, 64 gigs Ram, 100's more gigs of space on and SSD. You also need to download a couple of more required components. Those are "ae.safetenser and "t5xxl_fp16.safetensor". They act like the "VAE" layer in SDXL, but have to do with how the language model interacts with the image generator. The 2 I listed are the basic once that released with the dev version of Flux, but there are others out there. From what I read, they really aren't any better than the basic.

Good luck!

Flux/chroma will work if you have a solid 16 Vram and 32 gigs of ram, but it will be twice as slow as it could be, making the whole endeavor more frustrating.

File: S3_Hugo.png (636 KB, 417x1317)

636 KB PNG

>>3082047
Looks very nice. I know it wasn't the original intention, but he looks very similar to Hugo from Suikoden III.

File: 00001-4276090790.png (2.55 MB, 1296x1728)

2.55 MB PNG

>>3082125

100% WRONG SIR

I put that very picture of Hugo into https://imageprompt.org/image-to-prompt to form the basis of the positive prompt.

I was TOTALLY copying Hugo! lol

It would not get his hair or skin right enough for me, so I gave up. Will probably do a "native" type character soon though...

File: 2loras_test__00098_.png (1.43 MB, 1216x832)

1.43 MB PNG

Fuck, I want to breed Sokka deep

File: 00177-849000940.png (2.19 MB, 1296x1728)

2.19 MB PNG

>>3082157
Everyone loves Sokka!

File: 1753053455748923.png (149 KB, 500x455)

149 KB PNG

Any links to /ydg/ - Yaoi Diffusion General #6 it's now deleted. I was wanting the pics.

Need help fixing this prompt. I keep getting cat ears on the helmet and the horse emblem never shows up or looks right.

solo, 1male, human, sentai warrior, tokusatsu, dark pink suit, black accents, white accents, form-fitting suit, bulge, muscular male, athletic build, symmetrical body, sleek tokusatsu, helmet, heart-shaped belt buckle, horse shaped emblem on chest, confident stance, mouthless helmet, stylish costume, dynamic pose, heroic presence, tokusatsu hero,

File: 00107-2422477875.png (1.95 MB, 1296x1728)

1.95 MB PNG

>>3082212
Popped prompt into Homosimile V4.0, no Loras wat all. The odd helmet issue is baked into the "sentai/tokusatsu" prompt. You can't have one without the other. Putting "cat ears, ears on helmet" helped to change it to a helmet accent instead of ears (sometimes devil horns would pop up). I bet that the training data includes all the odd villains from those shows as well as their odd costumes. The fake "helmet mouth" is also baked in, though putting mouth in negative got rid of any seam or teeth like marks.

As for the Horse emblem, you are SOL. The matrix didn't understand that at all.

Positive Prompt: very awa, masterpiece, best quality, absurdres, newest, perfect quality, detailed background, intricate details, male focus, solo, 1boy, sentai, tokusatsu, dark pink suit, black accents, white accents, skin_tight suit, detailed bulge, toned male, narrow waist, heart-shaped belt buckle, horse emblem on chest, fighting stance, dark abstract background, spotlight

Negative Prompt: cat ears, ears on helmet, mouth, lowres, worst quality, bad quality, bad anatomy, signature, username, logo, watermark, artist name, old, bad hands

Used ReForge U.I.
Steps:50
Sampler:DDPM
Schedule type:SGM Uniform
CFG scale:7
Seed:2422477875
Size:864x1152
Model: homosimileXLPony_v40NAIXLEPSV11
Denoising strength:0.4
Hires CFG Scale:7
Hires upscale:1.5
Hires steps:35
Hires upscaler:R-ESRGAN 4x+ Anime6B

>>3082217
thank you

novelai is the only way to get consistent details + quality

maybe a chroma lora could help reach something similar, but every time I've used chroma by itself it's been pretty trash quality-wise

also, the creator is trying to make it work with qwen for better prompt understanding in the future

File: 902828333778128035.jpg (224 KB, 1152x768)

224 KB JPG

>>3082263
How did you manage to get the tattoo right? Sometimes I generate images of that characters and yet it always come out with the wrong tattoo no matter what I do use or input.

File: gggff.png (804 KB, 1294x797)

804 KB PNG

>>3082303
It's made with nanobanana, i'ts a bit censored but I made it resting with some random underwear i found in google and inpainted on top of them, You can do wild things with that model, i wish it was not censored.

File: 1.png (1.25 MB, 1120x928)

1.25 MB PNG

>>3082305

File: 1756495271471-b3a6ba2e-27(...).png (1.31 MB, 1024x1024)

1.31 MB PNG

>>3082306
This is Volt Krueger from the bouncer using just the picture of his 3d render and some pants I found in google, it's one of my teen game crushes

File: masterpiec-444636190-18_1(...).png (854 KB, 768x1152)

854 KB PNG

I did some stuff of the dad from Mirai (2018)

File: 00077-2074079132.png (2.33 MB, 1728x1296)

2.33 MB PNG

>>3082307
Kou from The Bouncer was always one of my all time favorite crushes! Really wish there was a good Lora of him. Anyway, have a Zach

File: 00128-2868891444.png (2.01 MB, 1296x1728)

2.01 MB PNG

Was messing around, was surprised how well Zach and Cloud were trained in, but I guess with there popularity it shouldn't have been.

File: 00078-1762908158.png (2.12 MB, 1728x1296)

2.12 MB PNG

SDXL

File: 00091-1762908159.png (2.26 MB, 1728x1296)

2.26 MB PNG

Was very easy to get them in bed together.

File: 00198-3222266151.png (2.28 MB, 1296x1728)

2.28 MB PNG

Hassaku XL v1.2-Style

>>3082332
>>3082331
Can you swap the roles, please?
>>3082336
Fucking nice.

>>3082305
Is that a collab or the Gemini one?
>>3082303
HOT. That is a daddy!

File: 00080-1762908157.png (2.2 MB, 1296x1728)

2.2 MB PNG

>>3082350

File: noir.jpg (278 KB, 1856x1280)

278 KB JPG

I blew up my week's worth of Civitai tokens in a single day, oops.
>>3082364
Aw yes.

>>3082307
>Volt Krueger
finally some love for him, I thought I was alone. and he might be the reason why Cid Highwind is on my list of crushes as well.

File: 00002-4322170317.png (1.78 MB, 1792x1280)

1.78 MB PNG

>>3081781
I'm training a lora for that artist right now if you wanna try one of the epochs: https://litter.catbox.moe/f74i10njaj2pgdmc.safetensors

picture is just the first one I generated with it

does these classic artists got their own lora?
https://exhentai.org/tag/artist:terujirou
https://exhentai.org/tag/group:neyukidou
https://exhentai.org/tag/artist:bulwyn
https://exhentai.org/tag/artist:matsu+takeshi
https://exhentai.org/tag/artist:takezamurai

File: tmp5ngggj_y.png (1.77 MB, 1280x1280)

1.77 MB PNG

>>3082400

They all have loras if you search on civitai. Terujirou and bulwyn probably don't need loras because they have enough art tagged on danbooru.

>>3081606
This is so hot

File: 903464555873623781.jpg (272 KB, 1024x1024)

272 KB JPG

>>3082306
I do apologize for my late response.
That it do look an interesting method. I will thank you for your kindness on showing how it is done. But may I ask, did you use the program in your computer or a collab/web version?

>>3082307
Volt Krueger is indeed a character that do deserve all the porn it can get. Such a rare specimen in a test-game Square Enix did for test the Playstation 2.

>>3082400
I do remember a lora of Bulwyn on civitai. But it was not for Ilu, only Pony.
Generally, a lot of what you asked are not in any page. And I will love if they existed for be legends of old.

>>3082303
I am testing even more with this character for create same effects in other characters that I want to experiment. As such, I will share some of the experiment, making him older and in to a full demon.

File: 903441294330750796.jpg (224 KB, 1152x768)

224 KB JPG

I will do appreciate if they are share on the internet.

>>3082426
>a lot of what you asked are not in any page...
it's over...

why not just make the loras you want