AI-generated posters don’t have to be horrible
john.hartnup.uk1526 points by ereiamjh a day ago
1526 points by ereiamjh a day ago
A lot of people in the comments say that the supposedly better examples in the article are still “horrible” because they can tell they’re AI and point to things that the AI did “wrong” and a human artist would do better.
That may be true if you’re working with a top of the line artist, but in my experience the average freelance designer online is significantly worse than AI. I’ve worked with dozens of budget friendly designers on Fiverr and consistently get way worse results than I’ve gotten from AI models. Will a great human designer produce better output than an AI? Maybe, but a small local event or business probably won’t have the budget to hire them or know where to find them. It’s cheaper, faster and easier to use AI to generate something that’s good enough for their purposes.
People also point out all the little flaws in the “good” examples and cite them as obvious tells that they were AI generated. I think it would be interesting to see if people could reliably distinguish between AI generated flyers and human made flyers. Sure, sometimes it’s obvious like when the AI adds those weird, cartoony people. But if the model is told not to create designs like that and stick to the styles used in the second half of the article, I expect people would have a much harder time distinguishing the human designs from the AI designs.
The biggest problem that I have with most AI posters is that they are too busy, you have to look for information instead of seeing it popping right away.
Amateurs posters of before AI didn't have this problem. White background, the important info in big letters, maybe one clipart or to. So I'm not talking about how pro artists could make a better job, but even the small local non-profit used to produce better posters. Not in the sense of "better looking", but in the sense that they were more readable.
At least the examples of this blog are more readable than what we usually see.
This is very true and after rendering 100 prompts twice for https://john.hartnup.uk/poster-prompts/ I realised that the prompts could be improved by asking for much less clutter. The best examples on there pick out one main visual subject, but many of them are too fussy.
I may do a V2 one of these days.
Clutter comes with the territory to some extent though. I was recently asked to do a real world PTA fayre poster, and the client (my wife) very firmly wanted 12+ attractions mentioned on the poster. Tombola, hook a duck, bar, BBQ, blah, blah, blah as well as date, location, sponsor, ...
To get a really clean design, you need to edit it down to the essentials. I don't know what pro designers do - convince the client to say less?
Unless the prompt says otherwise, AI will try to get a visual element in for every single word too.
Pro designers are generally paid because they’re extremely good editors. And they’re an intermediary between clients and their consumers.
So when you put the capability in the hands of the client without the editing (or taste) and no way to evaluate the result, you get the passable, homogenous thing.
It mostly doesn’t matter because there’s no economic incentive for it to matter, which is why much of our aesthetic world is the way it is.
Designers having that artistic taste, and clients lacking it, is exactly why I don't think generative AI is going to do much to replace (skilled) designers anytime soon. Likewise for other creative professions, and indeed for knowledge work in general.
Taste is a great solution for abundance, so I agree. Also, there's definitely a rubberband effect where products & designs need to differentiate on their design to stand out in a homogenous world, and you truly do need creative professionals having that judgment. Even using AI as a tool, you have to have the sayable vocabulary (at the very least) to agitate the aesthetic status quo.
And low-skill graphic design had already been mostly replaced by Canva, so I don't think AI is really having as much impact on the design industry as people might suppose.
Knowing what is in the realm of possibility can help you get much more out of genAI.
For example, you have to know about a style to know to ask for it or some variation on it.
I think we ALREADY at the point when AI is capable to replace like 99.7% of designers. It is only a question of market inertia and personal brands that prevent it from happening right now.
I think this generalizes to a lot of professional work.
LLMs are very good at satisfying the requirements you give them. They’re much worse at knowing which requirements should be challenged, reframed, or ignored because you don’t know enough about the field to know what actually matters. That’s why they’re such force multipliers for experts: the expert can spot when the model chose the wrong abstraction or checked every box and still produced something bad. A novice often can’t. I’d argue the model mostly raises the novice toward “average.”
“Tell it to push back” doesn’t really solve this either. Too little and you get “aye aye captain!” Too much and you get Opus 5 / claudise, where it seems compelled to find something wrong with everything.
I see this in a Home Assistant project I’ve mostly vibe coded. I can ask Codex to make the kitchen brighter on school days and it’ll happily implement something. But maybe an HA expert would say “you really shouldn’t model it this way.” Maybe they’d even question HA itself. The model could conceivably know that too, but it only sees a tiny slice of my world. There are infinite side quests it could raise, and a good professional’s real skill is knowing which one matters enough to interrupt you about and which 99% to silently ignore.
>It mostly doesn’t matter because there’s no economic incentive for it to matter
Very generous way to say "executive greed has drained bespoke design out of society. I've been seeing more videos on such a topic, but this one was by far the best one I've seen yet: https://youtu.be/HgjXjPP_p3s?is=yVsto8SCnY-SsnVY
(Just in case, don't let the thumbnail fool you. The car simply frames as it expands to general society).
I tried to give this video a good-faith watch, but found myself disagreeing with much of the evidence, conclusions, and research. It seems to be trying to diagnosing cleanslop — the minimalist and flat trend, which renders poorly as being clinical and lifeless. My point was that economics drive aesthetic, and in that framing, I think we both agree corporations will deliver the least amount of design tolerable.
Design-differentiated products still exist (Apple as mentioned, IKEA, Nike, Linear to name a few), but they're outnumbered in the current landscape and arguably still greedy.
>Design-differentiated products still exist (Apple as mentioned, IKEA, Nike, Linear to name a few), but they're outnumbered in the current landscape and arguably still greedy.
I believe that was ultimately what my response wanted to emphasize, and partially the call to action of the video. There is great design out there, but you need to wander through oceans of slop to find it. I had no clue about "Superblocks" until that video.
Likewise, to link the ideas back to Herr; AI is creating an ocean of slop that will make it nearly impossible to find the truly innovative used of AI out there. Curation was already a hard problem pre-AI and this only increases such difficulty by magnitudes. Among reducing said slop (which is nearly impossible now), the call to action is to have proper tastemakers lead the charge and bring out good ideas. Not just ones that generate the most money. Curation by definition almost always lowers view traffic to increase customer satisfaction, and that's not something our ad-based datascape wants.
That's a really backwards explanation for styles that executives commissioned, funded, and popularized in the first place
I think it's in fact forward thinking, in that modern executives lost the taste their older elites had. Or at least they fired the tastemakers for those who pleased them personally, to the sacrifice of proper design. Probably a bit of both, along with other factors.
The more charitable perspective (that falls more important point #2) is simply that taste is different now among the modern billionaire. Someone who can wander onto a stage in a turtleneck and denim jeans simply won't be the type of person who wants a "loud and bold" car. But they might still want to say that they had Ferrari make it. Jony Ive's perspective on design may simply be the product of the people and environments he surrounded himself with.
I can't remember where, but there was an article about the luxury homes billionaires are building themselves. They tent to have he design aesthetic of a bland, but expensive, corporate hotel.
As one of the AI-wielding amateurs, I got positive feedback from an audience member on this Ai-gen poster for a talk I gave earlier this summer https://drive.google.com/file/d/1MaawQcupC5XHqWBiTujSpKOKohx...
The clutter is there yes - it's not a sparse 1970s kitsch cultural event promo or 1970s sci-fi book cover (random example I found in 2 minutes https://www.etsy.com/listing/715136429/david-bowie-moonage-d... ) that's for sure - but I wouldn't want to remove any of mine's elements as each one informs the viewer at a glance their decision whether to attend or not.
I’ve been working on a thing where a bunch of raw content goes through a summarization pipeline and then gets manipulated with a bunch of checks to clean up copy and reduce clutter.
And then the visuals get added with a set of fixed templates. It’s not fully there but is reducing a LOT of the clutter.
Some of the results are even what I’d call good (and I’m a designer). I think a good harness could make non-cluttered results, if one of the AI labs was interested in a “less is more” approach.
Just taking one of your prompts and adding
> N.B. Make sure all the important information pops out (is visible at first glance) like an expert designer would, the information is more important than the design.
Already have me a much better result with the 'Arts & Crafts style', but N=1 so YMMV
Those posters all look pretty damn good to me. Thanks for that page. Super helpful for me and I need that right now.
There is very often a certain softness to the fonts in generated images. Why is that and is there a name for it?
Thinking more, I thinking the softness is actually inconsistency. Because the letters are generated, no two letters will be exact same and if zoomed in some edge might not even be perfectly straight. Another thing we can do these days (I think author should really try out in V2) is to use code to generate the . You may have seen examples of generating Javascript based videos with Claude. Same can be done with posters. Generate a font sheet and use that in code generated poster with image generated background.
Are these "fayre" events some kind of UK thing? I also didn't know offhand what "tombola" is and it comes across like it's part of a collection of standard activities for a "fayre" that have no cultural significance to me.
(Canadian here. Not everything is shared across the Commonwealth, I guess.)
Yes, fayres are popular in the UK especially in schools and small villages. A Tombola is a type of raffle where people buy tickets and the winning numbers are picked from a rotating drum. Normally done to raise small amounts of money for local causes.
Say you want to add an artist or change some information. Regeneration will end up with another poster altogether, right?
A new image at a technical level, but all modern autoregressive image models are skilled at only changing what is absolutely necessary if asked to make edits.
Well, low-stakes clients are usually poor clients because they’re clueless and overly demanding.
Real designers would avoid them in the first place.
Tangent I guess, it would be interesting to have like a AI/not AI quiz. They still look AI to me but idk if it's because I know
It should be clear in the text, disguising as non-AI is not a goal. It's allowed to look like AI. It's just trying not to look shit.
And that default look - that might not be shit either. Except we've now seen it repeated so many times that we're sick of it.
Am I biased to thinking it looks like shit because I know it's ai though and I think the AI look is overdone?
Suppose you have a budget of $20.
Do you think you will find good artists working for that fee?
I don't think so, unless they are working to prove a point -- that humans can be as good as AI.
And that's the problem:
Humans can be as good as AI, but they are much more expensive.
Maybe you have a friend that's trying to get started in design and could do a favor for a chance to build a portfolio.
Well what's better ? I have so many pictures of posters of old events or storefronts saved on my phone because they were so astronomically bad it'd have been better to not have any poster at all. Today's AI makes it look even worse but I'm hopeful that in five years we'll all live in a world generally more agreeable to look at and not "concrete with pink and yellow fluo paper smashed on top of it" like https://www.alamy.com/stock-photo/music-poster-wall.html or https://www.alamy.com/stock-photo/france-poster-wall.html?ps... because the cost of making things enjoyable with more colour and detail will have come down. Maybe we can finally go back to Art Nouveau if the cost of making it comes down!
For what it’s worth i think the Bi Nuu poster you share here as reference for something badly designed looks really cool.
Poster design like any visual or artistic field will always cater to conflicting individual tastes - for some it looks great, others hate it.
I like Art Nouveau but a lot of stuff in that style looks kitschy to me.
> we'll all live in a world generally more agreeable to look at and not "concrete with pink and yellow fluo paper smashed on top of it" like https://www.alamy.com/stock-photo/music-poster-wall.html or https://www.alamy.com/stock-photo/france-poster-wall.html?ps...
Sorry buf I'd take these posters over your “generally more agreeable ” homogeneous slop literally any day.
As I said, "Today's AI makes it look even worse". But the post makes me hopeful than in a couple years, actually good graphic design can come out from AI. It still looks terrible and AI even for the best posters it highlights, but noticeably less than the random ice cream or coffee shop where I live with fully ai-generated interior design, memorabilia and menus.
I’d rather look at the most amateurish indie posters plastered all over a wall than waste even a second looking at machine-generated nonsense.
> too busy, you have to look for information
Have you ever seen a bus schedule, or a parking sign, or a list of events at a local library or an “open hours” sign at a restaurant? Those are 99.9% unreadable, busy, and you have to look for information. And even then it’s not always clear if you got it right. And that was all before the AI.
There is a difference between busy for no reason and busy because of information density. The busy examples in this article aren’t information dense. They just have a ton of different flower types, random animals, background characters, unrelated structures, many different colors and many different shades of those colors, etc. But the information they are trying to communicate is minimal.
A bus schedule or a parking sign is very information dense. It’s trying to communicate a lot in a limited space and as a result require getting used to.
Those examples are filled with real info. The AI posters are filled with meaningless three word slogans, lists of meaningless copy and art where everything is labeled.
None of those are primarily marketing tools, they're things you use after you've already decided you want to use the product. They don't need to grab and hold attention and convince you that they have what you want, they just need to convey information.
Bus schedules are not a good comparison because they have to fit a ton of information into a single page. And I disagree that they are unreadable. Speaking as someone who commuted by bus for years in many different cities, they tend to be quite easy to read because the tabular layouts make them very easy to scan quickly. Once your eyes anchor on the page, you can find what you're looking for without effort.
The biggest problem in my view is: It's not editable by hand. There is no production data I could open in InDesign, Illustrator or Photoshop (Layers) and change things to my or my clients liking.
I am not familiar with the field, but if there isn't a way to have LLM's generate graphics as editable InDesign or Illustrator files (maybe something Adobe will sell you?), I am sure there will be soon.
This is a problem which is entirely solvable by existing and new design/production apps. Instead of working entirely with raster image agents should programmatically edit the target file format.
You could try having, say Astra, use one of these programs via computer use to design it. I haven't tried it, but I hear it is really good at computer use. Might even have more of a dialogue as it creates, so dictating choices at each stage. Would take longer, but does not require someone who has expertise in the tool, just good design notion.
That also reminds me of AI-generated text, which is also (when done without caring) barely readable because of the useless noise mixed the relevant information that a human that proofread the text would have thrown out.
People act like making a poster is this absurd Sisyphean task that takes a degree in art to produce.
You don’t need to hire an artist, you don’t need AI. Just download Affinity for free, find a nice font and color palate, and put text on the page. I’ve done my own event posters and they take less than an hour to produce with next to zero prior experience and look massively more appealing than AI slop.
There's an amazing viral twitter thread where someone claimed a real Monet painting was an AI generated imitation, and it got an avalanche of comments from people who claimed it was an obvious fake and a poor imitation at that.
It's not tracking a real thing as much as people would like to think.
This is exactly the problem. Most of the time when people criticise the output of genAI, what they’re actually commenting on is the ethics of AI rather than the output of it. And they then attempt to use the demonstration as a proof of their point rather than forming an unbiased opinion of the output based on its own merit.
Now I’m not saying there isn’t legitimate criticisms that can be made about the ethics of AI, I’m highly concerned about the problems that AI brings too. But that’s a separate argument entirely.
It’s a little like people who are pompous about expensive wine, whisky’s, etc and then fail blind taste tests (plenty of studies there too).
I've generally been very impressed with genAI artwork in "impressionist" styles. But thinking about how the diffusion models work, that seems like exactly what you'd expect them to be good at. Things that call for perfectly straight lines and constrained patterns (and explicitly symbolic elements, like text) seem to be considerably harder but apparently very doable now.
There is still obvious genAI slop being put out there all the time, and you can tell it's genAI slop because it's slop: even doing something at that level by hand would be more time and effort than anyone would bother putting in unless it's paid work that then justifies getting it right and doing it professionally. But now that the sloption exists, there's a flood of people taking it.
Prose, on the other hand, is another story. It became possible to produce "grammatically correct, academic-sounding and generally using words to mean what they actually mean" very quickly, but years later it's still rarely if ever good, it just goes through varying different quirks. Maybe part of this is that the training data can't properly reflect what the target audience was for various pieces of text. A lot of it, surely, is that many people are easily impressed by "grammatically correct, academic-sounding and generally using words to mean what they actually mean", and don't themselves have much of a sense for actually good writing.
I don't think this is true. There are people who claim everything is AI – maybe even to bolster their ego.
But for a lot of things I can with high probability discern between AI and human in every day appearance and I know many other people can as well. One of the things I look for is specifically certain types of extreme... not just flaws... just overall atrocity. When I read posts submitted to HN I often realize that something is bad writing before I can specify this to AI text. Pangram almost always agrees with me. Here is a recent example: https://news.ycombinator.com/item?id=49715657
I don't doubt that there are certain scoped tasks that certain models are quite good at. But in my every day life I constantly encounter AI slop that is so bad that I take it as a personal insult that someone thought it would be okay to subject me to it; to presume that I am so tasteless and numb and have no quality standard whatsoever.
I mean just look at the example posters at the beginning of the article. "The problem with these is not so much that they’re bad." Yes, yes it is. They are horrible. I cannot understand how anyone can describe them as anything but that.
> But for a lot of things I can with high probability discern between AI and human in every day appearance and I know many other people can as well.
It’s easy to say that when you already know the answer. But my point is people frequently misattribute stuff as being AI generated when it wasn’t. The GPs Twitter example is proof of this. And you only have to look at the number of times people accuse someone of being AI on HN because they happen to use em dash or an Oxford comma (or any other writing style that was part of the corpus models were trained on), and it’s pretty clear that “obvious tells” are just guesswork.
There was a study years ago where a group of individuals were given white wine with red food colouring, and they were asked to describe the wine. All the participants described plumb and other notes that are exclusive to red wines.
These weren’t dumb people. And the people who claim to spot “obvious AI” aren’t dumb either. It’s just very easy to identify tells if you believe you already know the answer. Unfortunately that’s just how fucked up our human psychology is.
> One of the things I look for is specifically certain types of extreme... not just flaws... just overall atrocity.
So, please forgive me, but this means that you are good at finding out some patterns in AI output. There are always a lot of false positive and negatives, and I haven’t found a convincing study showing that people are good at finding AI in general.
What helps is that the market is very concentrated so if you know GPT-isms and Claude-isms that’s basically the current situation covered. It does not mean that you can reliably identify AI in general.
How do you know you can discern AI vs human with high probability? You need some external source of truth to judge that, otherwise all you know is that you can discern bad, obvious AI. See also https://en.wiktionary.org/wiki/toupee_fallacy
I can say that my false positive rate is very low which is enough in this context.
I’d expect false positives to be low. Bad AI has a distinctive style and is easy to pick up on. Just like bad toupees are easily recognized. The false negative rate would be far more interesting.
The false negative rate would be interesting but is irrelevant in this context. I was replying to a comment claiming that most people take issue with the ethics of GenAI rather than the quality and that people have a high false positive rate. I raise myself and my experience with people I know as a counter example.
There might be AI generated things in my every day life that I do not recognize as AI, but that's not what's being discussed.
How did you check? What was your oracle when you thought something was AI and when you did not?
I mean, this situation is a textbook case of ground for a lot of common fallacies, starting with confirmation bias. We need to remember that our brain wants to be tricked. If you think you are special, you are not. You’re just not aware of it and even more prone to falling into these traps.
How do you know that?