Thread #43171882
File: 1760571150158079.png (2.9 MB)
2.9 MB PNG
/aia/ - AI Art Thread #77
>Mentally ill artist discovers a way to express themselves through AI art
▶WARNING
>By posting in these threads, you recognize that AI generation is a valid artistic process.
Please respect AI artists like any other artist.
Backends:
>https://github.com/LykosAI/StabilityMatrix
This is more of a one-click manager that you can install / run the other backends though. Easiest by far. Recommended to download Forge through it and use its built-in Civitai browser to download models with.
For models not through the in-built downloader just download and drop them in Stability_Matrix\Data\Models\* (lora folder if a lora, TextualInversion if embedding, stablediffusion if checkpoint...)
>reForge
https://github.com/Panchovix/stable-diffusion-webui-reForge
I would recommend installing this through Stability Matrix.
>ComfyUI
https://github.com/comfyanonymous/ComfyUI?tab=readme-ov-file#installin g
Has a learning curve but is extremely customizable and usually has the latest methods / papers implemented first.
>"But I don't have a decent GPU"
NovelAI also does great pony: https://novelai.net
You could also use a service like civitai.
Models:
>Pony Diffusion V6 XL
https://civitai.com/models/257749/pony-diffusion-v6-xl?modelVersionId= 290740
>NoobAI-XL Great new alternative. Is closer to novelai than anything else. Really diverse yet coherent positioning / styles, knows artist tags.
https://civitai.com/models/833294/noobai-xl-nai-xl
>ZoinksNoob, a popular new model built off ChromaXL Mix, which itself was built off of NoobAI. Very, very close to NovelAI and arguably better at many things, it does ponies very well and has much less of a "burnt-in" default style than something like Pony Diffusion V6 does.
https://huggingface.co/zatochu/ZoinksNoob
>Purplesmart.ai best for starters and with a great community
https://purplesmart.ai
>Useful artist style LoRA:
https://civitai.com/models/317578/pdv6xl-artist-tags?modelVersionId=35 6175
>LoRAs by /mlp/:
https://rentry.org/ponyxl_loras
>LoRAs by /h and other useful info:
https://rentry.org/ponyxl_loras_n_stuff
>How to make your own LoRA:
https://rentry.org/59xed3
https://civitai.com/articles/91/how-to-correctly-obtain-images-for-a-d ataset
https://civitai.com/articles/3522/vaIstrixs-crash-course-guide-to-lora -and-lycoris-training
https://civitai.com/articles/4100/how-to-create-a-style-lycoris-for-po nyxl-on-low-vram-locally
>Prompt tips for new users:
Grab an image you like the style of from the desired model's civitai's site and drop it onto the PNG info tab in SD, you can then click send to text to image to reuse its generation info for your own images. Change the seed so you do not get just the same image. Use tags from e621 and natural language.
For the love of CelestAI, DO NOT FEED THE TROLLS
>Previous Thread
>>43113320
153 RepliesView Thread
>>
>>
File: 1768397452100037.png (2.7 MB)
2.7 MB PNG
>>43171891
WDYM?
The OP pic took about 15 minutes of iteration and the only fix was to remove a phantom signature from the bottom right with a single inpaint pass.
>>
>>
>>
File: 1744550677847158.gif (285.5 KB)
285.5 KB GIF
>>43171911
>1girl complainer
>>
File: 1761904538460639.png (2.7 MB)
2.7 MB PNG
>>43171909
Fuck if I know. Probably around 100 000 or something.
>>
>>
>>43171927
>15 x 100,000 = 1,500,000 minutes
>1,500,000 minutes = 25,000 hours
>25,000 hours = 1041.67 days
>1041.67 days = 2.85 years
You've spent nearly three years genning when you could have spent three years actually arting.
>>
I'm at like 10 000 should pump up those numbers and give the slop complainers something to really complain about. But alas I can't be assed. Saving up for a 5090 desktop, will have one by this time next year.
>>43171937
>says the guy that doesn't utilize the newest tool in the artist's toolbox.
>Because?
>Just because?
>OKAY?!
Lmao, get replaced.
>>
>>
>>
File: 1775160903798509.png (706.6 KB)
706.6 KB PNG
>>43171938
It's about 20 seconds per raw image. Vast majority of the raw images are looked at once and then deleted whenever I purge my folder.
I only have about 100 or so images that I'd consider finished, and half of those are now so old I won't post them without touching them up with the new tools that have been invented since.
>>43171947
You're trying to compete with a machine. The machine does it faster and cheaper and is much easier to learn to use. And also keeps getting better in ways a human never can as the tech improves.
Sure, it's still lacking in many ways.
>>
>>43171953
>It's about 20 seconds per raw image
>Vast majority of the raw images are looked at once and then deleted
What is vast majority? Is 15 minutes a short time or a long time in terms of achieving a "finalized" image?
>>
>>
>>43171947
>>43171950
get replaced
>>
>>43171953
>You're trying to compete with a machine.
Did the Lord say that machines, ought to take the place of livin'?
>The machine does it faster
Ok
>and cheaper
Electricity and a computer are cheaper than pen and paper?
>and is much easier to learn to use.
Skill issue, unironically.
>And also keeps getting better in ways a human never can as the tech improves.
Right, and by "better" you mean "approaching human ability"?
>Sure, it's still lacking in many ways.
At least you admit it. In what ways do you think it's lacking?
>>
>>
>>43171972
>>43171965
>rawwwwwwk
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>43172005
And people that speak in "le" unironically are slop humans. Get replaced.
>>43172007
>:^)
>>
>>
>>
>>43171972
>>43172015
>Not recognizing Johnny Cash
Children should be banned from the internet
>>
>>
>>
File: 1762204031072698.png (2.2 MB)
2.2 MB PNG
>>43171956
I started this one shortly after you posted. 15 minutes is about as short as it gets, unless you go for something really really simple like a popular character solo without a specific pose in mind. For this one, I ended up making a total of about a 100 images to reach pic related. I'm not entirely sure about the total since a lot of the time I watch the generator work live and abort the run if it doesn't go in the direction I want.
There really is no upper limit to how long it can take. That depends on the complexity and how close to your vision you want it to be. Adding more characters vastly increases the time spent since the gen can do duos somewhat passably but anything beyond that is a total crapshoot and you're better off genning them separately and stitching together the final pic from pieces.
>>
>>
>>43172042
Don't you have anything better to do with your life than obsessively troll the thread about the thing you hate?
>I deliberately entered a thread full of things I hate and now I'm mad! How could this be happening to me?
>>
>>
File: where do you think we are.png (2.4 MB)
2.4 MB PNG
>>43172048
>Don't you have anything better to do with your life
pic related
>obsessively troll the thread about the thing you hate?
I don't hate AI. I use it pretty often myself (goonbots), and I just want to understand the AI image genner's perspective.
>>
File: 1771912583450533.jpg (10.9 KB)
10.9 KB JPG
Reminder
>>
>>
>>
>>
>>
>>43172074
>implying
>>43172067
was saved from here since I thought it was relevant to this thread and the amount of hate it gets.
>>43172080
lmao
>>
>>
>>43172086
no that's what's his face, the one that draws down syndrome faced ponies.
>>43172089
>>43172013
this one. lol
>>
>>
>>
>>
File: 328794327942.png (80.6 KB)
80.6 KB PNG
>>43172101
>>43172039
>I started this one shortly after you posted. 15 minutes is about as short as it gets
>still riddled with aislop errors
>proompter is just too low iq to know how to fix it,
>Will never develop a critical eye throwing more slop at the wall
>And too lazy to fix it further
>still claims it's an artist
>>
>>
>>
>>
>>
>>
>>
>>
File: 1708689758569.png (1.3 MB)
1.3 MB PNG
>>43171891
Perceived laziness. Been ITT since the start and I think what gets lost in these discussions is how messy and indirect this process actually is.
And honestly that’s exactly where the frustration comes from.
With a *skilled* artist, the pipeline is straightforward and under control. They think of something, and they execute it step by step:
>Sketch for structure (perspective, anatomy, pose)
>Refine lines and style
>Apply color intentionally
>Shade and render
>Fix mistakes exactly where they happen
At every stage, they know what they’re doing and why.
With AI, it’s completely different. What I do looks like control on the surface, but it really isn’t.
My actual workflow is more like:
>Start with an idea in my head something very specific
>Try to translate that into a prompt (already a problem, because language doesn’t map cleanly to visuals)
>Pick a model (PD, ZoinksNoob), maybe a LoRA, tweak weights, CFG, steps, sampler, seed…
>Hit generate and wait for a complete surprise
And then the result comes out… and it’s almost right, but not really:
>Pose is off
>Expression is wrong and bedroom eyes in a sad situation is cringe.
>Composition drifted
>Go back
>Rewrite parts of the prompt
The worst part is models usually are good, but they have their own bias.They want to generate certain faces, certain styles, certain compositions. Not fully in control I think I'm constantly fighting the model’s preferences. It feels less like creating and more like negotiating:
>Trying to describe an image to something that doesn’t really understand you, and it keeps answering with "close enough" variations that miss the point.
Meanwhile, a traditional artist just draws the thing and will have completed the artwork just to brag about hos good artists are. They're not wrong. That’s the part that stings. And after all that effort I end up with something I had to fight to get. Then I post it, gain a few likes maybe 20, 30 on Tanta but next to it, someone posts a simple goofy drawing and it gets twice the engagement and doesn't stop.
And yeah that’s where I had to confront something myself, overestimating my output. Thinking this is technically better, so it should perform better, and then being confused when it didn’t and that’s where the Dunning–Kruger effect hits. Perception vs reality.
So yes AI art is a valid process. It takes effort, iteration, and problem-solving but pretending it’s easier, faster, or more controlled than drawing?
I would say an artist have the imagination, the imagination is an IMAGE in the brain and the hand is the printer but let's skip the "just grab a pencil" lie. not everyone can physically handle a stylus right like they can and directly output and move the stylus on the screen. I feel more like a studio photographer finding a place, a spot, asking a model who only speak Lithuanian and a bit of Ukrainian to take various poses and expressions.
>>
>>43172741
Lol it doesn't sting at all, I'm getting the best coom that the legacy artists never managed to make for me on this exceptional quality level and at zero cost too. I don't give a flying fuck about whether any of my gens would fetch 100, 10 or even just 1 upvote, I gen for one person only: me, myself and I, don't even bother uploading any of it especially not for ungrateful fucktards.
>>
>>43172753
Yeah, and that’s fair. I used to think the same way.
But over time I realized I don’t just want to be someone who generates images, I want to aim higher as an artist. I want to build a name, have some identity behind what I make, maybe even change a few people’s minds about AI art instead of just adding to the pile. Because right now, something feels off and the system is rigged.
I make content I genuinely like. I spend time on it, refine it, iterate on it, try to push it beyond "just another gen." But building a fanbase from scratch, without connections or visibility, *is* hard. Way harder than people admit.
And at some point you start realizing: it’s not just about quality. It’s about who made it.
Let’s be honest the same image posted by an established artist will get attention, while yours just sinks. Not because it’s worse, but because there’s no name attached to it. No recognition, no built-in audience, nothing to anchor it.
That’s why a lot of people end up gaming the system. Buying followers, pumping up the numbers, just to look established enough to be taken seriously or to shill commissions. It’s not even about ego at that point, it’s survival in an algorithm-driven space.
And even that isn’t reliable anymore. I lost hundreds of followers recently to a purge, and it just reinforces how artificial the whole thing is. Those numbers don’t mean loyalty, they don’t mean engagement they’re just a fragile illusion of visibility. The reality is brutal: if you’re under ~1k followers, you’re basically invisible. The algorithm doesn’t push you, people don’t notice you, and your work no matter how much effort you put into it just gets buried. While some absolute moron with 1200 followers post a stick figure "LOL NAZI HORS" and gets instant hundred likes. So yeah, I’m trying to move past just “making images” and figure out how to actually exist as an artist in this space. Because right now it feels like you can do everything right and still be treated like you don’t exist. And I feel this is going to be our biggest problem, to pierce through a noise of slop and people who claim drawn art is more organic or some shit.
>>
>>
>>43172067
>Why putting transgender and disabled people in the same bucket.
I am legally disabled and I could not express myself without AI. AI is a godsend for millions of disabled not a whining minority of people who pretend to have disability.
Like them or not but AI devs aren't your MAGA memer guy from Twitter and clearly not working at X, but the blue haired furry Anon makes fun of. If you cared a little about asking and engaging with people who create your tools.
Yikes! the requirement is being a decent human being!
>>
>>43172822
>>Why putting transgender and disabled people in the same bucket.
Because they're the same thing. Admittedly I have slightly more respect for actually physically disabled people rather than those who disable themselves for virtue points.
>>
>>43172741
>>43172798
You use too many ChatGPT-isms, lay off the chatbots for a while, son. And you seem to believe that traditional art is easier and has a lower bar of quality for popularity, yet you autistically stick to doing things in a way you know firsthand doesn't get you what you want, which is clearly attention and recognition. Is it just sunk cost at this point?
>>43172822
>I could not express myself without AI
And that's a good thing
>>
File: WhiteLightningNAI_e000014_00_20250712195833_2.png (1011.4 KB)
1011.4 KB PNG
>>43172824
>those who disable themselves for virtue points.
Wow fucking really? This is such a terminally online take it hurts. You’re comparing actual disability, something people live with, suffer from, and navigate every day to identities you don’t understand, and then flattening it all into attention seeking. That’s not insight, that’s just you being unable to process anything outside your own frame.
From where I stand (enby and AI artist), this just reads like insecurity. You see people modifying identity, presentation, or even the tools they use (AI, etc.), and instead of asking why, you default to fake or self-inflicted. That’s lazy thinking or confusing,
Also the irony is insane: you’re on 4chan, a place literally built on anonymity, roleplay, and identity fluidity, trying to lecture others about what’s real.
And this whole trans / enby / disabled people hate AI the most thing is just straight up BS as anon mentioned. It’s the opposite in a lot of cases.
AI art is accessible. That matters even if it has a strange learning curve and even if it isn't as efficient or lacks the social recognition artists have, you don’t need years of training. You don’t need perfect motor control and this is an essential point for AI, I could prompt with just my voice. You don’t need expensive education or connections with others to know what to do.
For a lot of LGBTQ who are statistically more likely to be marginalized or excluded from traditional pipelines, AI becomes a way to create anyway. To express identity, experiment with presentation, build avatars, tell stories. Yes we exist and yes, I feel more at ease with the kind of anti AI hate from my moots, where a right wing AI artist will get at minima being told to kill himself. I see this everyday in Discord channels or X and I feel bad for these artists, I should not but it's a fact.
Same for disabled people; If drawing physically is difficult or impossible, generating is literally a workaround. A tool that removes barriers instead of reinforcing them. That’s not disabling yourself, that’s removing limitations.
Think of a smarter reply next time, it's a it insulting.
>>
File: 1768659223262437.png (2.2 MB)
2.2 MB PNG
Anyways, here's boop.
>>
>>
File: Gemini_Generated_Image_knedbkknedbkkned.jpg (2.7 MB)
2.7 MB JPG
>>43171615
RAINBOW trout
>>43171535
I already did that one, Anon:
https://tantabus.ai/images/67985
https://tantabus.ai/images/67984
https://tantabus.ai/images/67983
>>
>>
File: Gemini_Generated_Image_871cnf871cnf871c.jpg (2.4 MB)
2.4 MB JPG
>>43173546
Bonus image of MY BABY so she can have more images on Tantabus than any other OC
>>
File: screenshot.png (102.1 KB)
102.1 KB PNG
>>43173557
BOOM
>>
>>
>>
>>43172039
This is the promoter process. It's time to evolve.
The Ai artist process is
>know what tags you want for exactly what you are thinking (filly Luna corrupted by eldritch tentacle beasts only takes like 20 pos and neg tags once you know what ones to use their actually function)
>generate 4 and always have at least one good one
>photoshop and Inpaint the fuck out of it
>post to social media for 100 likes and be depressed.
>generate furry slop
>inpaint for 20 seconds cause you don't really give a shit about furry
>get 1000 likes on Instagram and 2 patreon subs
>cover the pain by genning more ponies.
>>
>>43173668
People think its spammy here and tantabus because it looks like basic 1024 dalle prompts from 2023 while most of the ai art fandom has moved on to illustrious, yiffymix, and other models in a variety of resolutions and styles.
You picked the most dalle style you could find
>>
>>
>>43173712
That's basic feedback. Nothing trolling about it. Surely you can see posting dalle style over and over with bizaare ai prompts like riding fish is going to annoy people.
Go pull up dalle threads from back then. That's what these images are. Nothing wrong if you like it but don't expect everyone to think it's brilliant.
>>
>>
File: 30 seconds in stable diffusion.jpg (3 MB)
3 MB JPG
>>43173786
stop being a faggot and have some fanart
>>
File: dfj3owp-ae9955ec-3b87-464e-a6bc-36d07c1d0f6a.jpg (948.3 KB)
948.3 KB JPG
>>43171882
I've been seeing some vore art of this character, does she actually eat people?
>>
File: 1771023861629354.png (3 MB)
3 MB PNG
>>43173690
Anon that is my process. Out of the 100 pics, only like 20 were raw gens and the rest were messing around in inpaint. And the floaty Twi had more raws than usual because I had trouble getting her to actually float without her having a human pose.
And I share your pain. Ponies are, honestly, a fucking pain in the ass to work with because they're so far removed from human/anthro anatomy that the gens do best. And then there's the fact that while I do like shit like pic rel and spent hours upon hours wrangling with tags and weights to create the style, posts that are just "1girl, ass, pussy, from behind" get like 5 times the attention and popularity because of coomers.
Tho I don't have a patreon even if I would probs get a fair bit of coombux form there if I did.
>>
>>
>>
>>43173806
No. Her entire characterization in the cartoon is just "a supportive loli". If you see her being or doing anything weird, interesting or memorable in derivatives, that's on the derivative's author(s), for better or worse.
>>
>>
>>
>>
>>
>>
>>43171891
I am actually giving up, and switching back to my graphic tablet. Fed up with this bullshit. Years ago, when pdv6 got hyped I thought we would get something bigger, like whole episodes generated with og voices and all that but everything is paywalled and is not getting cheaper.
>>
>>
>>43174526
Good. More content for the AI to use. I would not completely abandon prompting. But keeping eye skills is probably one of the most important parts of being a prompter. Some of the things I've seen published show that many prompters don't even look at their final product before posting it. Or is possible they do but have no eye. Creating art from tablets will make your eye strong so that when you do decide to prompt you'll know what to look for, correct, reject, etc.
We didn't get to where we are by people giving up. Advancements have stagnated but that doesn't mean breakthroughs are impossible. I don't think AI can produce it's own breakthroughs. I think that is where artists and creators step in.
>but i don't want AI to get good and make great art available to everyone
Someone is gonna do it anyway. Maybe you can make your mark on it.
>>
>>
>>43172863
> I feel more at ease with the kind of anti AI hate from my moots, where a right wing AI artist will get at minima being told to kill himself
This. Leftist AI hate - I don't like it.
Rightoid AI hate - You are an affront to God you're brown you're not working hard, a communist tool just good for generating memes, don't you agree Grok?
This puts things in perspective. I hate slop and the kind of repetitive shit like
>>43173542
>>43173546
>>43173794
flooding, but I don't want this anon to kill themselves.
>>
>>
File: 1757658863907494.png (2.8 MB)
2.8 MB PNG
>If Celestia had been born in a different place and time.
>>43174526
You've been saying the exact same thing in every thread for more than a year now.
How's it going with the drawings? Anything to show?
>>
>>
File: Gemini_Generated_Image_v439ilv439ilv439.jpg (1.9 MB)
1.9 MB JPG
Incredible hihg speed
>>43173794
>short mane
>suspiciously hidden cutie mark
See, this is the problem with trying to use other models for this
She cute though
>>
>>
File: 1749378718305966.webm (1.9 MB)
1.9 MB WEBM
I have a 3D print I am printing at the moment that's based off a picture that was posted in last thread. It will be ready in 20 hours. It was of this. If anyone wants the file to print out for themselves, I can upload it after I can confirm it works. Or if you just want it for the 3d model that's cool too. Ill post a picture of the finished product tomorrow around noon.
>>
>>
>>
>>
>>
>>
>>
Its going much faster then what the slicer said it would take. It's going to be finished in 4 hours. It's already doubled in size from what I just posted. I love this printer. I don't even remember the settings I used. I just threw it on the SD card after slicing and hoped the settings would work. It's been like 3 years since I have played with it.
>*fingers crossed*
>>
File: 💀🌙🦄🎞️.webm (3.9 MB)
3.9 MB WEBM
>>
>>
>>43176239
>>43176549
Updating in /create/ at this point, I kind of branched off into another thread and I don't want to cross streams. Ill post the finished model here when it is done.
>>
>>43174533
I literally have no idea what you are talking about and I don't know who that pony is made after. I just thought it was a robo-ponywaifu kind of deal. I am glad to see my choice garnered some content. It was just the best looking one for an AI thread IMO. No more, no less.
>>
>>
File: 1771345042642124.png (1.6 MB)
1.6 MB PNG
They glow in the dark.
>>
File: 1747314009871717.png (3.5 MB)
3.5 MB PNG
I think I broke something.
>>
File: Gemini_Generated_Image_xvcjaqxvcjaqxvcj.jpg (2.5 MB)
2.5 MB JPG
Moth
>>43176737
She zoom
>>
File: Gemini_Generated_Image_qzlameqzlameqzla.jpg (2.5 MB)
2.5 MB JPG
>>43177777wew
>>
>>
File: 💀🌕🦄🎞️.webm (3.9 MB)
3.9 MB WEBM
AI Pony in meatspace.
>>
>>
>>
File: chicken and scootaloo.png (969.1 KB)
969.1 KB PNG
working on a new project while the old one prints out another to paint. This one still needs a bit of refining for a print ready model.
>>
>>
>>
>>
>>
>>
>>43178922
I could, link me a bunch of AI Derpy imagesto make and I will decide which one would work for me and I will give it a try. Mainly I am looking for pictures that would have her model balanced sitting on a desk, basically no floating or flying. I mean I could try one later but it's just easier to display on my rotating base thing if it sits flat. Maybe sometime later I will try floating/flying ones, it's just I don't have a stand for it. I guess I could print one though...but anyways enough rambling. That's my only requirement at the moment.
>>
File: 1749913972126260.png (884.8 KB)
884.8 KB PNG
>>43178935
NTA but here's a Derp for you.
>>
File: Gemini_Generated_Image_h3zrclh3zrclh3zr.jpg (2.3 MB)
2.3 MB JPG
>>43178436
RAINBOW waterfall
>>
>>
>>
File: hercules olympus.png (2 MB)
2 MB PNG
Anyone would be willing to take on request of making some picturesque gens of Cloudsdale architecture (preferably with closer birds eye view on the urban Roman/Greek style buildings being made out of clouds)?
>>
File: 1770645025529776.png (771.4 KB)
771.4 KB PNG
>>43178978
Modeled it and I am printing it now. The supports are a little sus but we will see in 24 hours.
>>
>>
>>43179459
>>43178978
It's still printing but here's a preview of the model
https://u.pone.rs/dnfrmkpy.webm
>>
>>
File: Willem_Dafoe_Looking_Up.jpg (46.7 KB)
46.7 KB JPG
>Mentally ill artist discovers a way to express themselves through AI art
>>
>>
>>
>>43177803
>>43179699
Neat!
The head doesn't look too good in 3D space but I think that's more a general problem of ponies not translating into 3D at all since a shape that looks like they do in the show cannot exist in three dimensional space.
>>
>>43180504
Yea I get where you are coming from. But considering I was trying to create a non-Euclid form in real life I found it quite acceptable.Maybe the back of her head is to puffy but thats just her hair volume. And her big brain. Also she has some eyesockets that could probably use some smoothing out. But like you said trying to create 2D in 3D is neigh impossible to get right at all angels.
>>
>>
>>
File: Gemini_Generated_Image_5lir2s5lir2s5lir.jpg (2.4 MB)
2.4 MB JPG
Fireflies
>>
>>
>>
>>
>>43181034
>>43181035
Keep going, champ. These images are cute.
>>
File: 1773852676488143.webm (3.9 MB)
3.9 MB WEBM
>>43178978
It's done printing, I messed up on the tail so I can't remove the supports without breaking it off
>>43178922
It is done
>>
>>
>>
>>43182459
You know you could probably make a lot of cash making these custom sculptures for MLP fans, especially if you also paint them. Might be a lot of work if you're not in need of the money though. Anyway, first genuinely interesting use of AI technology I have seen yet (since you actually also put some work in it).