secret horses manifesto
made in like 4 hours on an empty stomach for the manifesto jam 2026, using my love letter engine. photos and drawings by me. all instances of synthetic content have been omitted.
html transcript on my website.
stuff mentioned:
Development log
- secret horses jam37 days ago
Comments
Log in with itch.io to leave a comment.
this is actually smth i think and talk about a lot, so it's very fun to see someone who has similar thoughts as I do... I disagree with how you've put some of this, I think, but fundamentally I do agree with you.
i miss how old AI generated art used to look a LOT. i think what you said about wanting to see people use the medium in ways other than to "maximize anime women output"... pretty much exactly sums my feelings on it up.I MISS "name one thing in this image" SO BAD... i loved the fucked up computer pattern recognition and output. i miss her. humans definitely CAN hit that level of abstraction, but the way human and machine pattern recognition work ARE inherently different, and its so intereseting... idk. i miss hher...
i don't know what you know about vocal synth? i'm gonna assume not a ton. but there's a lot of commercial AI-trained singing synthesizers out there — but they're ethically trained, and the community generally understands this, so everyone's chill with it. I think that in most cases you Could probably get around the AI asset ban as long as you communicate that your assets are locally + ethically trained? i think it's bizarre that I have yet to see that "as long as its ethical and not environmentally harmful" sentiment echoed elsewhere, but i DO suspect it exists. ...it's just that genAI's primary usages have understandable bred a LOT of contempt for it.
i will say: i pretty strongly disagree what you've said about using AI to generate assets to quickly create something that feels polished. I think, honestly, that this echoes part of the issue I have with most genAI output right now: we're defining "polish" in a decidedly bizarre way here. as the toxic yuri jam said: it doesn't need to be "good"! if you can code, you can open MSpaint and scribble an idea with your mouse real quick! and you CAN design a game in such a way that this CAN be a final asset, and it CAN be effective. i want to see more of that! because the way someone makes that asset, even if it's a 30 second scribble, still tells you so much about how they see the world. i too like when people have the tools, but it they DO exist outside of AI already. you can make a game with final sprites that are all photographs of pen scribbles on the back of receipts if that's all you have. the true metric of polish is simply the cohesion of the vision...
...sorry, this got. quite long. I trimmed it down, even, but as I said: i have a lot of feelings and thoughts about this. thank you for making this and sharing yours.
i love getting long comments, don’t worry at all! thanks for taking the time to reply in such depth!
i’m not really familiar with vsynth, that’s cool to know! and yeah, i think it’s weird you don’t really see this nuanced approach to ai. of course i understand a blanket ban is a lot simpler to enforce, but i think the alternative is worth exploring. that’s partly what i’m doing with the jam (sorry for self-plugging hehe but it’s something i’m quite excited about…)
funny that you disagree with what i said about polish, because i 100% agree with your thoughts! polish is a vague word that doesn’t really mean much, does it? i think it’s overall / in a vacuum pretty cool that one can create something that matches their vision. this is something that i’m not sure people really understand when they criticize the idea of AI “democratizing art”. because while it is true that anyone can make art, broadly speaking, many people do not have the technical skills to create what they want to create. however, this is where you are absolutely correct: it’s important to question why one would want or feel the need to create something that aligns with very clean, “professional” / commercial work in the first place. there’s a very narrow definition of what constitutes Good art / game assets, and that’s fucked. sukeban games talks about this in the worms blog post (that’s a great vn too, btw). it can be much more interesting to work with these limitations, and it often makes for more unique works of art.
When I was a kid I would see stuff about AI generation, like old CaryKH videos and tweets from the person who wrote "AI generated Batman" (which in retrospect wasn't AI but I was insistent it was lol) and thought it was so cool and interesting. This manifesto really touches on that possibility space that enthralled child me, but I wasn't able to understand (partially due to not having public image models back then).
Don't know if I will actually use AI in my projects, as I've come to find I love creating digital collages myself, and despise prompting on a visceral level (in large part due to how pushed on everyone The Sludge is), but I would be lying if I said that AI mush wasn't an inspiration for the surrealness I try and go for, and it seems like a fun space to explore. :>
oh wow i bet the earlier generative images were particularly fascinating for a child… they’re truly unlike anything else you really see.
i personally also quite dislike prompting… something i really love about using ganbreeder / artbreeder is the more exploratory approach to generation. there’s no real way to get quite what you want, it’s a process almost closer to automatism in some ways…
Oo interesting, maybe I'll check it out then.
I was mainly referring to text based stuff since AI image models weren't really a thing back then, but I did also find non-AI generative art super fascinating too. They were very similar in my head, and still are to an extent, just using different technologies.
true
incidentally I recognised secret horses at a glance and your redraw says something to my soul
I look forward to the day that this whole AI bubble finally pops and there's just the soap scum floating around amidst the cultural debris. There, in that cultural froth perhaps we will see a renaissance of the algorithmically-derived weirdness you and I and most of the other commenters here yearn for...
one of the early synthetic image projects i really loved was mushy. i saw it back before i ever made or shared anything online, so i didn't "have a reason" to download it and never did (foolish), but i remember how much i dreamed about all things i could make with it. the crunchy/vague textures are soooo beautiful to me and it bums me out that the itch page was taken down. now the only place you can find the credits for the work it was trained on is the wayback machine!
omg yes! i hadn’t thought about it for ages, the noisy fuzziness is so pretty. i really wanted to use it too but didn’t know for what, i think tile based things just don’t come naturally to me. this makes me wonder about using ganbreeder stuff with wareware…
Thanks for making this!! It's honestly so sad that the pre-AI-boom generated art got to be such a casualty of the (justifiably angry) movement against the AI boom. I've seen some events receive criticism for not having clearer or stricter rules against AI, and honestly why do people have to be such cops? We're missing out on good things, and you're right to point it out.
Interesting read, I think it's necessary to consider this. It depends on how it's used, but still I fundamentally disagree and think it's nonsense. Don't really see the value in it.
RIP to @ai_metal_bot, your images used to be so interesting before the AI boom
yeah.. im not a user and i dont think i have ever used any of the old generators and all that but like.. i sorta hate how like everything just falls under "ai" bc i actually love the idea of like using ur own art to have a machine make stuff with it.. kinda the whole thing with like an avatar generator when u set it to random and it picks stuff from the assets it has. i think curation has a lot to do this it.. and purpose and intent.. like a friend had a twitter account that was a bunch of stuff they wrote but like it chose how to mash stuff up from the database or whatever.. how is that really that diff from "ai" writing when "ai" is taking writing from the internet/etc that other people wrote and mashing it up? its just hateful that "ai" is being added to like everything where its making shit worse.. and trying to replace people.. meh its complex
as ai hype starts to die down a bit, i feel that some space lets up for conversations like this. well, maybe it's wrong to say ai hype is dying down. maybe we're all just getting used to talking and thinking about it.
> since when do we all believe in the sanctity of copyright and in the immortal human soul?
it is so unfortunate how "ai" use has been pushed by corporations as a means of cutting costs and accelerating productivity. and im happy that our creative communities have taken such a strong stance against that. at the same time there are genuine ways of artistic exploration using these tools and methods, ones that dont benefit these corps. so yes i agree!
#yup
wouldn't it be crazy if there was another jam playing with the artistic intent of genAI
may we set our bones down instead of picking them in the future with our friends
ugh i’ve really been thinking about hosting a jam or something…
i found this one very quaint. the way the meta theme creates the “blogging about blogging” effect. from a very different game-making bubble for sure.
An interesting perspective, I mostly agree with you. I don't like genAI, or rather what it's become synonymous with... The fast, cheap, cookie-cutter, everything-looks-the-same-now way of making things really appeals to people with no artistic sensibilities and that don't want to pay for assets. I also think a lot of (commonly chronically online) artists pay far too much attention to the «content theft» side of things (which I couldn't care less about) rather than the real issues with the practice, such as the mentioned environmental impact.
I'd be lying if I said AI-generated images/videos of old weren't deeply fascinating to me (nowadays they're commonly referred to as AI hallucinations, which I find to be simply inaccurate as machines aren't sentient, and even if they were how would they be able to perceive the world the same way we do?). But, the community surrounding these things aren't artists nor even art appreciators.
I once learned of an incredibly skilled artist that was once creating his own AI image generator from scratch based on his own images (I believe it was a glitch artist called Rui Martins, here's his Tumblr). My dad, a former graphic designer, had a lot of fun creating things with legacy Midjourney. Then the culture surrounding AI changed because it became mainstream.
I hope I'm making some sort of sense. Without being too much of a hipster, I feel that the only way to make generative AI not suck is to make it uncool, so I'm inclined to side with the «no AI» rule.
as a certified AI hater, I actually mostly agree with all of this. I do see artistic merit in exploring the latent space of machine learning, combing through the machine's hallucinations to find something strangely beautiful.
but I don't think there's anyway around the fact that if you don't put these disclaimers about no AI, you're not really going to get a lot of that... you're going to get sludge. you're going to get people who see the craft of creation as an obstacle rather than the point using AI to pump out art where all the decisions were made by a machine whose entire purpose is to create a randomized average of art.
I'm no copyright supporter by any means, but it's hard to ignore the malice in big tech's use of artists' data--they don't seek to help us, only to replace us.
I would like to go back to the days of ganbreeder--I played with it a lot myself back in the day, and it was really fun!--but I don't know how we can do that without opening the door to the right wing art haters. I hate to cede ground to them, but there's no way around the fact that they have claimed the territory.
I'm not sure how to reclaim machine learning as a genuine artistic tool without supporting their agenda when the people making the largest and most prominent models are right wingers themselves.
i absolutely agree that noAI tagging effectively functions as an anti-shovelware filter nowadays. to me it’s kinda like asset flipping, in that undoubtedly there are devs making good work that’s only possible through pre-made and stock assets, even though the overwhelming majority isn’t. how do you separate chaff from wheat? i don’t think it’s possible in a large platform: passing judgement on each individual work is unfeasible. maybe with smaller collectives, webrings and the like.
ultimately i think the way to not be lumped with the lame mainstream right-wing crowd is to just make work that’s obviously ideologically different. it’s not really a matter of Reclaiming AI into a leftist technology through putting together the greatest talent show this town has ever seen, but just like: i’m so deeply uninterested in the themes and aesthetics of right-wing creative production that i just have to put faith in my audience to understand what i’m doing, to see the existence of an artistic vision, the use of the medium as a deliberate choice. from the reception of my work, i believe i’ve been successful, even if i do still feel nervous at times.
i think it’s reductive to say the medium is unsalvageable because of the ideology of the corporations / startups in the space. in fact i believe openai and the like benefit from the perception that they are synonymous with AI and that it is and can only be a service provided by big corporations. when in reality people have been running open source models themselves and training their own stuff etc. perception has never really been about truth, though, which is why it’s complicated.
That’s very fair. I agree that there are small local models that don’t have the baggage of supporting Big Tech. I guess we just have to be really clear about the distinction between the open source stuff that they aren’t profiting from and the big shiny models that they are.