No, You Don't Need AI Art, Not Even "As Placeholder" (Indie Games, Taste, and Slop Aesthetics) (Patreon)

Hi folks.

I've been having a lot of AI Artpocalypse kinds of thoughts recently. So I thought I'd post an update today to let people know where I stand on things. This will start with some personal perspectives, move on to some very practical "why I think it's a bad idea to do x" type gamedev stuff, and end on some reflections on what this means in terms of community.

I'm now past the point where I feel like I can really trust any information about the environmental impact of sticking a bazillion GPUs in a warehouse. People seem to also generally not care that much if chatbots are creating new forms of psychosis which would've seemed hamfisted if described in an 80s cyberpunk novel.

Intro: "Mass Appeal" is Bad Data for Indie Media

Even if we could magic away all of the moral issues with AI technology, I contend that most of the outputs are crap. A lot of this comes down "this looks bad and you only can't tell because you have no taste". I think in a lot of cases this is still true. And in some cases, it doesn't matter. If you want to slap bad AI art on a poster for a small local business I think the amount of people who will be bothered by it is pretty negligible.

But when people use AI art as part of small "indie media" productions, I think it betrays a misunderstanding of why people choose small indie media in the first place. It makes people feel like they've gone to a specialist bistro and been served a microwave ready meal. And the problem is, as someone operating in a niche indie media space, you're much, much more likely to be serving to people who know and care.

The way AI draws and writes is kind of crap because it's an approximation of "good" based on statistical majorities. "What do the largest amount of people like?" can be helpful contextual information if you're making Call of Duty and Candy Crush and your goal is moving the needle 1% on a bunch of KPIs. But as an indie creator, the mass audience isn't your audience. It never will be, and you shouldn't want it to be, either. Small artisinal eateries are not competitors to McDonalds.

If you don't see a problem with using statistical majorities to decide what's aesthetically "good", then let me regale you with an insight from one of the great philosophers of our time:

I am sure that there are some very skilled writers, artists and musicians who are able to incorporate gen-AI into their workflows in ways which create a good or interesting result.

I do not deny the possibility of this. But the inevitability of a "good" possible use of a thing does not preclude it being a "bad" thing in general. Like heroin. Or Coldplay.

The problem is that even if the basic output gets "better" over time (as we're told, constantly, is inevitable)... even then, the criteria for what "good" output is will continue to be tuned to appeal to the tastes of the lowest-common-denominator consumer - i.e. making things which are overly polished, glossy, "poppy".

This will always lead to outputs feel "wrong" in the context on indie media, where the "scrappiness" and jank is part of the charm.

To use a different medium as an example; AI music is never really going to "threaten" underground punk or hiphop, because those are genres where authenticity matters, and where that authenticity is directly linked to the aesthetics of how the music looks and sounds; i.e. made cheaply and with the tools available to poor people.

On the other hand, AI music will threaten more polished, commercial forms of popular music, like trance, radio pop, K-Pop, softer and more prosaic singer-songwriter acoustic etc. precisely because the audiences for those types of music were never interested in the notion that the music was an expression of the performer's authentic experience and creative craft.

Here is how I suspect AI "creativity" will actually impact videogames.

Using it will become the norm in large polished productions (many will argue it already has, but I'm not so convinced). Most of the audience pushback will most negatively affect small indie developers who decide to use the technology where it's most fervently unwanted. And this is because indie game consumers care how their games are made more than gamers who exclusively play AAA and mobile f2p games.

If you're someone who'd love to be able to use the technology unfettered, this might seem unfair, but it's not exactly a new pattern either. People hold small mom-and-pop businesses to a higher standard than McDonalds or Walmart, and that's because they're easy to personify as ethical agents doing good or bad things in the world (even if the real-world impact of a Walmart or a RockStar Games is far larger than the impact of a Toby Fox or a small local pizzeria).

You'll notice that, so far, I haven't talked about AI coding agents. That's because, by and large, I'm completely ignorant about them. I don't want to use them because I enjoy solving problems myself or with a collaborator.

I try to be agnostic about tools like Claude, because I've seen good folks making decent stuff using it.

Conversely, I've yet to see anything involving visual art or creative writing which didn't make me think "I'd prefer to see what you came up with yourself". And this is because, whenever anyone proudly shows me some piece of game art or writing they used GenAI to produce, it... doesn't look indie.

Which, I know might sound like a snobbish cop-out, but I really think ties into some broader problems lots of devs have with thinking indie game dev is just make a small crappy fan-homage to a AAA game you like, rather than a scene with a specific set of common design patterns and aesthetic associations.

So, here's a deep dive into some of my current thinking on AI aesthetics, leading up to why I'd rather you just didn't join this Patreon/Discord if you insist on letting a machine be in charge of how your games look...

Problem 1: You Cannot Statistically Model Yourself Out of Tackiness

Putting aside me being a massive hipster asshole for a moment, let's look at this problem on a purely commercial level. The issue with using AI generated audiovisual outputs for indie games, is that they tend to lean towards over-polished aesthetics which don't fit the rest of the project, and are also unlikely to "land" with the sorts of people who buy lots of smaller games.

This is noticeable with music outputs which feel too "expensive" for the game they've been put into, and writing which is overly flowery but has no "voice", but most noticeably with game art which looks like it belongs on some AppStore game about popping bubbles, and using microtransactions to pop them slightly quicker.

"The home-made "jank" is part of the appeal. If you want indie games to look like AAA games, you're probably not a fan of indie games... which means you're probably not very suited to making games which appeal to indie audiences."

This is something you can already see happening on Steam, where a bunch of capsule art pieces look more like something you'd see advertising a casual game for a phone, not something someone would spend $10-$20 for. It's not so much that the outputs are "bad", it's that they are "too good" (i.e. polished) in the wrong direction.

If a developer is predominantly a techy person who doesn't have much artistic knowledge, or a strong sense of how art direction maps onto tastes and preferences, you can end up with a false sense of security, because you think "polished and professional" is all that matters, when it really ends up sending unhelpful signals. Take the following example:

Let's say I'm the ideal audience for Camp Manager Simulator (shown above) and I have no idea that the above image is AI (or I simply don't care). The glossiness of it sends particular signals, namely that this is a large-scale, commercial project made by a team of 30-100 people. If the "glossiness" of the game fails to match up to what's presented here, it's going to be a big shock, and that's when the bad reviews start.

As an indie developer, your art should be as janky as the worst part of your game. Anything else is setting you up for failure. It's better to have "bad art" which fits the polish of the game, because this means players show up with the correct expectations. Bad reviews often stem from having packaging which grossly misrepresents the product.

Coincidentally, this is also why I don't make my YouTube thumbnails look snazzier than the videos they represent. This is basic "4 Ps" marketing theory stuff. People who deliberately buy and consume homemade, DIY media do not want it misrepresented to them as glossy, mass-market media. And people who are sold something as if it's glossy, mass-market media are more likely to be angry if what is inside the packaging is actually homemade, DIY media.

The Camp Manager capsule example is AI generated. I know this, because the developer admitted it on Reddit. But you'd have a similar problem if you hired a capsule artist who was very technically competent, but didn't understand audiences and aesthetics (plenty of those out there too).

Problem 2: Isolationism Robs You of Opportunities

So, you'd be right here if you said something like "so, the issue isn't actually the AI, it's that people don't know how to prompt it correctly?".

Well, yes and no.

The problem I see, regardless of how this technology "progresses" is that it's fundamentally not designed to ever push back on your ideas. That means the process of using it robs the operator of opportunities to develop taste and style.

We develop our taste by trying things out, by engaging with art and design which already exists and discussing it with people. The way gen AIs are currently designed to work robs you of these opportunities, because of their innate sycophancy and inability to tell you when something is a bad idea.

When you collaborate with a human, there's a much higher chance that they'll tell you when your ideas are bad. Someone who has created hundreds of game assets, or dozens of Steam capsules, or hours of game music, is going to probably have a better sense of how aesthetics map onto audiences.

This is why £50 of freelancer time - spent on the right person - will always be a better use than £50 of generator credits (not to mention the fact that you get another organic node in your project's marketing network from that person posting about the game wherever they can).

When you collaborate with a human - whether as a paid freelancer or just a friend you're jamming with - you get their taste as well. If you're a solo developer who has made it far enough to be thinking about things like steam capsules, survivorship bias usually dictates that you are probably someone who leans more to the technical side. With a human collaborator, you get someone who quite frankly has better taste and judgement than you in certain areas. With a bot, you get a sycophant.

And when your judgement in an area is weak, so you don't really know what looks cool, the last thing you need is sycophancy.

Problem 3: We Have Programmer Art For a Reason; (And That Ain't It)

The other common argument I've seen, lots of times, is that the AI art is just being used for "placeholder". This is usually an indicator that the person speaking doesn't really understand what placeholder art is, or what its purpose is in the production of a game, or any other media project. So let's spend a moment on that now, and after that please never fucking tell me you've "used AI for placeholder" ever again, thanks.

Placeholder art is, like any other component of a prototype, there to help you answer a question. When you place placeholder art in a game prototype, you're using it to answer questions such as;

How big should art assets be in our game, and how will this affect interactivity in terms of visibility, collisions etc.

What is our game's colour palette, and how do different objects need to be lit and shaded to stand out (or be hidden)?

What shape language should our game objects use to help communicate what type of object each sprite/model represents?

Placeholder art goes into a game prototype to help answer these types of questions, and to put something in the game so that the game can be played, tested, and validated.

It can also be helpful to hold back on certain art direction decisions until you know how the game feels. Perhaps you test different types of combat and find a more ridiculous "slapstick" approach just feels like a more interesting game than your originally planned realistic approach. In an ideal world, your art direction will be able to move with the project. But maybe not if you've already gotten too attached to something which prematurely indulges your fantasy of being the CEO of Blizzard.

For some great examples of what placeholder art actually looks like in finished games, check out this Kotaku article.

The problems with prematurely using polished, final-looking assets in a game are multiple.

Even if you plan to remove the AI placeholders later, your home-made assets will look "lower quality". This will make you resistant to making the change and finding your own art style.

The AI placeholders will be at a level of quality/polish which doesn't match other elements of your game (either visual, or in terms of how the game feels to play). See the previously stated principle: your art should be as janky as the jankiest part of your game.

If you do take your project to a point where you want to expand it or get help from a real artist, they are not going to trust you because you've put the AI art in the game. They'd be right to suspect that you might mix AI art in with what they've made, giving them a bad name, and they might also be put off by it because it sets a standard which is "too high" in some ways.

I like to put some "could pass as final" assets into my games early on, so that I can work out what my target art style is, but I never use other people's assets if they aren't in a style I'd be able to achieve on my own. This is my golden rule for using asset packs, and it should apply to using assets generated in any other way too; if you couldn't do it yourself, don't use it, because otherwise you'll be screwed when you need to expand the game and make matching things from scratch.

Having too much "could pass as final" art in an early prototype also betrays an underlying belief that the game is already a good idea and deserves to make it to launch. That's totally fine if you're doing it as a learning project, but it's not good practice for serious commercial game dev. A prototype tests an idea to see if it's worth pursuing. It isn't just draft 1 of something you've already committed to. Dressing up draft 1 with loads of over-polished crap is just going to make you more attached to your idea and reluctant to kill if.

What you should be doing is finding reasons to kill the idea, and then continuing with it if and when it survives the kill tests. Excessive decoration can be fun - and can serve other purposes if you're doing it as a way of actually learning how to make assets in a new style, yourself - but putting in over-polished assets from elsewhere is, in 99% of cases, going to be turd-polishing.

Problem 4: The Technology Attracts Antisocial People (Who Don't Actually Like Indie Games or Care About Other Developers)

A few months ago I made the active decision to stop answering questions about AI in streams. The folks asking them weren't regular chatters, and seemed to be people who were jumping around between gamedev spaces trying to find someone with a grain of authority to tell them it's ok, do whatever you want!

I've tried to avoid discussing the usual moral qualms about the environment or theft and meet people where they are on it. Some people are excited that they can now do something they'd always just dreamed of! I don't feel super enthusiastic about shitting on people for that excitement.

But at the same time, nothing I do here really works if people aren't coming to it with a willingness to learn, and I still think an hour spent making shit art is better for you than 4 hours prompting.

These aren't technologies which helps you to do something, it does it for you. Image generation is not making it easier to make art, it's replicating the experience of being an art director with a minion who doesn't say "no" to anything.

It also seems to attract people who - as I alluded to early on - are so attached to glossy Blizzard-meets-MCU high-polish aesthetics that they don't actually like any indie games. The sort of people who want to use AI to generate their game's art tend to think anything which looks a bit home-made, shabby, or naive is bad because it doesn't fit with their very narrow idea of how "professional" looks or feels.

These are the same sorts of people who will brazenly and without shame tell a community of 400 indie game devs that they "don't really play or like small indie games". Well then, maybe you might want to consider going away?

These people do not make good community members, because they don't respect the craft of others, nor find it interesting to learn about. They aren't interested in game development beyond the project they're working on.

These are the videogame equivalent of delusional smalltown rockstars who go to an open mic just to perform their own work, then sneak off without listening to anyone else's songs. Sneak being an appropriate term, because they know that participation in the community is meant to be reciprocal; they just don't care.

This is ultimately what's happening when you make the argument "I'm only using it as placeholder" or "well, I need to use it, otherwise I won't be able to make what I want to make!" It reveals that you don't actually like indie games. You just want to live out an unrealistic fantasy of making what looks like a $5mil AAA game on your own.

The problem I am having, generally, is that this technology - particularly in its audience facing forms (generating images or text) feels antithetical to the communitarian approach I would like people to adopt when making and marketing games (at least within the Indie Game Clinic community).

It now goes further than this was trained on stolen work or you should've paid someone to do this. The thing which bothers me the most, personally, is the way the technology encourages people to assume it's better just not to ask for help... or to do a bit of extra work learning how to do something (from people). It is a socially isolating technology, which preys on people's social insecurities by promising to remove the need to interact with humans... Humans who might say "erm, no" or "that's an interesting idea".

The saying goes "the cost of community is inconvenience". The takeaway was not meant to be "avoid community, gain convenience".

I understand the appeal. Learning is hard. Asking people for help is scary. Every time you ask another person to explain something to you, or playtest your game, or do you a favor, you're facing rejection. But these are all the sorts of things you are meant to struggle with as part of becoming a functioning adult and member of society. They are not temporary nuisances to be shoo'd away using a animatronic golden idol in the shape of Jar Jar Binks.

The underlying appeal of this technology, in a lot of cases, seems to lie in enabling people to avoid the discomfort of asking for favours. This is because, like I said, it's not really showing you how to do the thing yourself, but in the vast majority of cases simply doing it for you. It's giving you an experience of art direction, not of art-making.

Conflict avoidance and rejection sensitivity are two things commonly experienced by sensitive, creative types, and in my adult life I have spent quite a lot of time in therapy and reading self-help books to try to combat these traits in myself. The idea that you can simply avoid the discomfort of asking for a favour which might be rejected by getting a machine to do something for you instead is incredibly seductive.

And the big concern I have about this, from a community POV, is that communities which are too openly blasse about generative AI use will therefore end up becoming home to the most maladjusted and antisocial people.

It maybe not be an inherently antisocial technology, but the strongest appeal it has works along those lines.

A "Community" Is Not Just a Collection of Self-Interested Individuals

When I started the YouTube channel, I did so because I was sick of working on games I didn't really believe in. In my time as a professional game designer, producer, and creative lead, I've experienced extremes from studios where everyone had a lot of hands-on work to do, and a lot of ownership over the stuff they made, to larger, more structured companies which were very boring to work at because you were such a small cog in a large machine, even at senior levels.

I strongly support indie game developers because I believe that independences leads to better, more interesting games. As I've said many, many times on my streams and videos, I am not interested in helping every single person who makes games to "follow their dreams" or some similar nonsense. Indie games aren't just "whatever videogames happen when someone makes something without publisher funding". They're a unique subculture within games; a hidden garden where the cool, experimental, expressive, janky but authentic stuff grows and thrives.

Hollywood Weed Unc Seth Rogan was recently quoted as saying people who use AI to "write" shouldn't be writers and should "go do something else". He followed this up by saying that he doesn't want tools which make him write less, and that they're not appealing because he enjoys writing and it's that enjoyment which makes me identify as a writer. I feel very similarly about the automation of any sort of creativity.

Sure, there might be some parts of a process you would like to get rid of, and AI might provide some way of doing that without paying money to a human. And while I acknowledge that the Eric Barone "program the game, make the art, make the music" type of solo developer is not a realistic goal for all of us, I still believe it's a better thing to aspire to, compared to being someone who wants to work alone but doesn't acknowledge that DIY jank is more creatively interesting and fulfilling than automated slop.

I set up CollabJam last year precisely because I felt like a lot of the discourse around solo development on YouTube and Reddit was inherently antisocial. In some cases this is driven by a macho "I can do it all myself" kind of attitude. At the best of times, that's admirable, if a bit foolish. But it's also very easily coopted by people who want to sell you a $1000 course.

It's fine to want to do things on your own, but if that's the case, why not see making a game as an exciting opportunity to learn technology and art? Rather than viewing either of those things as some tiresome chore which you'd love someone else to do if only collaboration didn't mean compromising your brilliant, amazing vision of making Jar Jar Binks.

I set up Indie Game Clinic because I wanted to find a way that I could help people who were already developing cool games to make them even better. To put it in the most clinical business terms, I wanted the channel to be a "business funnel" so I could do "designing consulting", and gradually get to be pickier and pickier about the kind of projects I got to help on, until I never had to work on another free-to-play mobile game or fake piece of educational software again.

I didn't deliberately brand the channel as INDIE Game Clinic so that I could provide "business solutions" to people who don't actually like or play indie games... or people who like the idea of running a games business more than they actually like the craft which goes into making things.

I'm part of a generation of devs who started making games as a hobby long before the idea of making a living from it seemed even slightly plausible. Like Seth Rogan, I find the idea of trying to automate away the activity of actually crafting things a bit obscene. It's not so much that I don't think that it can't, someday, produce good or interesting outputs.

I never expected that the audience for what I make as a gamedev YouTuber would include people who don't really enjoy making games and simply want a shortcut to success. But the prevailing culture I entered into, at least on the YouTube platform, seemed to be one dominated by folks who see themselves as temporarily embarrassed millionaires looking for quick hacks, rather than folks who genuinely care about making cool stuff.

I make things because I enjoy it. The ambition of finding any sort of financial success doing something creative (whether making games, music, videos, or some mix of the three) was always a means-to-an-end; if the thing is financially successful, I'll get to do more of it!

But I feel like "I want a black box to do it for me" mentality is so alien to my own, that I don't really want to socialize with or support people who think that way.

I don't care at all if you use Claude, as long as you're not asserting that I should.

I don't even care that much if your art or music was AI generated (although I'll still tell you when it looks, feels, or sounds like shit, because that's literally my job).

But the thing which bothers me most about AI, at the moment, is the emboldening effect it's having on people who, in many cases, should be in therapy dealing with the fact that they can't play well with others.

I don't want to talk to gen-AI advocates, let alone enter into some sort of facade of "debate" with them. It's not because you're fundamentally wrong about every single thing. Maybe the technology is inevitable. Maybe I won't be able to tell the difference in five years time.

Mostly it's just because they're usually cringe.

As folks on Bluesky recently said in relation to Richard Dawkins' recent descent into full-blown bot-shagging, "this technology is designed for guys who think the stripper actually likes them".

Wanting sycophantic technology which gives you exactly what you want is cringe.

Being antisocial about the value of others' input and perspectives is cringe.

Being lazy is cringe.

Acting like other people are stupid for not being as "efficient" as you (lazy) is particularly cringe.

So I'm going to need to find a way of constantly and consistently telling them to bugger off, without becoming like a broken record. I'm still working on it.

Thanks for reading. and thanks to everyone in the community who, so far, has made it a wonderful place, and one which is capable of discussing the nuances of tools like Claude... without being a bell-end about how everyone should efficiency-max their gamedev by hammering slop out into the Steam-o-sphere.

On an unrelated side note, I have turned down somewhere between £10,000-£20,000 in sponsorship opportunities for companies like Unity AI Assistant Bezi over the past year or so.

If you're currently a free subscriber and would like to support the IGC community while also helping me fix the hole burnt in my pocket by my silly little outdated ideals, you can get 15% off an annual subscription by using the discount code BOTSHAGGERBEGONE before the end of May.