ereiamjh

Ironically, Comic Sans is more effort than Times New Roman.

But people would shit on a Comic Sans poster more than a Times New Roman poster.

Spring/Summer/Christmas "Fayres" are just a British thing. I'm not sure whether it's emerged as a spelling in recent years, or whether it's always been the case. But churches, village halls, schools, they've been doing it for ages.

That's not something AI has invented.

This is actually quite hard to achieve, at least in a one-shot LLM prompt.

You could engineer it by asking for a list of subjects sorted by obviousness, then pass that list to a script that discards the top 30% and randomly picks from the rest.

Even if you prompt an LLM to pick one at random, it usually won't, without an external randomness tool like that - which is curious, isn't it, when non-deterministic-ness is so inherent to LLMs.

V2 of the 100 poster prompts site is now up, and it has indeed stopped putting icons next to every single thing.

The moral is that with AI you get what you ask for. What you don't specify, it does however it likes.

I think a thought pattern that happens here is "Would you prefer a chair mass produced in a factory, or would you prefer a hand-crafted chair?"

Where by analogy, AI is the factory chair, human-designed is the hand-crafted chair.

And your mind imagines the most beautiful hardwood chair, with perfectly mitred joints, smoothed and waxed, your ideal handmade chair, made by a master craftsman.

Not the wobbly ugly piece of crap I would make if I tried to make a chair.

Ikea's cheapest factory-produced chair is a nicer chair than the one I would make by hand.

I think Canva Pro might do this - I dunno, I'm not paying for it.

You can certainly ask ChatGPT for a poster, then ask it for a collection of assets with transparency, so you can composite them as you wish.

It does take away from the "smash out a poster in 10 minutes" that a harried PTA volunteer wants, though.

I suspect the tool to do this as a one-stop shop is coming very soon, though.

I'm fascinating by how image models draw text.

I've a designer friend who's very distracted by the text on even a text-only gpt-image generated poster, because it looks like a font he knows, but differs in subtle ways.

I'm thinking, good, it's escaping from the tyranny of fonts. A font is limited by what a font layout engine can do. A human drawing letter shapes can decide that one 'f' ought to look different from another 'f' in the same layout, if that works for them. And so can AI.

This clearly works in styles like "60s psychedelic" where it's all swirly "hand drawn" letters. But I see no reason why it shouldn't work for "type". Perhaps bulk text would be more readable if the engine had that kind of freedom? If only we knew how?

The current generation of image generating AI probably isn't doing as good a job as it possibly could right now, but with the right training, all sorts of things are possible.

I'm from Britain, where in the 50s you might well have had an outdoor toilet, and still had food rationing.

BTW one of my fondest memories of poster-making in my 90s student days, was when we recreated the packaging design of Asda tortilla chips, for a Mexican night at the student bar.

Tequila slammers £1 :)

From the site:

"I recognise that almost all of the styles are entirely inappropriate for a village fête. The point is to demonstrate what you can do for any event, and to show that you can ask for whatever style you want."

Village fete is just lorem ipsum for the prompts.

Yeah, and I have opinions about [3].

I think pretending you didn't use AI, when you did, is unworthy. So I'm not trying to make posters that fool you into thinking it's handmade. I'm trying to make AI posters that aren't crappy.

I've got some AI writing too, and I don't try to remove all the clues that tip you off it came out of an LLM. I only try to remove the tics that make it read badly. And I present the work upfront as being AI-assisted.

The only problem there is, in the real world, there's a cohort of people who don't think "that looks crappy because it's AI" so much as "I don't care if it looks great; I object to AI on principle; I now hate the thing that poster is advertising". If you cared about that cohort, I guess you'd either lie about using AI, or simply not use AI. But I kinda wonder how long they'll still be around.

Mexican Calavera, you picked up on - might be one of the weakest, as it seems to apply a Day of the Dead layer on top of "default ChatGPT crappy cartoons".

It's not that CRTs were "forgiving", it's that designers leaned into what they'd do to pixels. A classic example is the tanks in Atari 2600 Combat - they look ridiculous on a hi-res screen. On a CRT they sort of look like tanks.

I have a JAMMA cabinet at home. It has a CRT in it. I usually play on x-in-1 pirate emulator board. So I don't mind that not all the hardware is authentic. But it's important to me that the screen is authentic, or it all just feels wrong. The curve must be there, the scanlines must be there.

I really get quite cross when you visit a museum and they've got retro games playing on flatscreen LCDs. Doubly so when it's meant to be a vector game.

https://news.ycombinator.com/item?id=49764791#49767382 had a lot of healthy discussion, and many people pointed out what I already knew, the posters were too cluttered.

Claude steered me into improving the prompts with actual evaluations... which cost me actual money...

Well, all the HN chat put me in the mood, so I did V2.

There is a lot more empty space, and lot less clutter, and a lot less over-literal interpretation of event details (just because a gig is at a venue called The Anchor, doesn't mean the poster needs a picture of an anchor)

That's a more interesting question than you might have intended it to be.

A designer's job has two contradictory aspects. One is knowing the rules and following the rules. The other is having novel ideas, and breaking the rules when it works.

The balance between those depends on the brief, and the designer's place in the ecosystem. For example the architect I paid to design my extension wasn't in the business having novel ideas - he designs well proportioned buildings in established styles, that work as living spaces, don't fall down, and get approved by planning authorities.

But other architects are paid to design skyscrapers that look like nothing you've seen before.

For a school fete poster, you're probably more in the first category - follow the rules. But you want to know where there's room for variety within those rules. Which is where the blog post starts - not every Spring fete poster has to have bunting and daffodils.

Tombola: ahead of the event you ask for donations of prizes. We do it for a PTA, so it's a "no school uniform if you bring a Tombola prize" thing. Sweets, booze, trinkets, etc.

Then you get two matching sets of raffle tickets. You put all of one set folded up in a drum. You take perhaps the ones ending in 0 or 5 from the other set, and tape them to prizes.

Then you sell 5 draws for £1, if you draw 5 or 0 you win the corresponding prize.

Very popular, big money spinner, legal without tedious admin.

A common meme is that a lot of the prizes get regifted time and time again, especially unwanted weird booze - the strawberry liqueur someone aunt brought home from Spain.

You're going back one generation, and I'm thinking further back.

Like, if we all settled for a 1950s British living standard, using current manufacturing technology, we could achieve that while everyone works a little bit less.

But, we'd prefer to keep labouring and have nicer things.

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.

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.

A lot of things are like that though. There's a cartoon - I thought XKCD but I can't find it - about how a beginner might identify an oak leaf by a set of descriptions - but an experienced arboriculturist would just say "well, you know, it's an oaky shape".

Like, I can tell a crow from a blackbird. Could I explain to you how to tell a crow from a blackbird? Absolutely not.

Thanks for sharing this @Bluestein - this is my site.

I'm aware of many of the flaws in these examples, but they're one shots and a serious user should refine and iterate.

The main common flaw is that almost all the posters are too busy. As well as wanting to fill space with detail, the AI is eager to represent every word in the prompt. So if it's a village fête for the church repair fund held at the cricket club, it wants to show a village with a church and a cricket match underway.

I'm considering doing a V2 that addresses these. It's a bit laborious as I haven't paid for API access to ChatGPT - I just copy-paste into the web app.

AI will leave blank space if its prompt gives it permission: https://john.hartnup.uk/poster-prompts/minimalism-art/

And similarly, all your other points. But of course, the more I say "you've got to prompt it right", the less value the AI is providing. A great AI would of course have that taste on your behalf.

This is something I actually cover in the follow-up project's how-to. Ask the AI to sense check the info you've supplied. It can tell you "hey, maybe you should mention venue address".

https://john.hartnup.uk/poster-prompts/how-to-use/

But as with all of it, it's no use if you don't ask the right questions.

(I'm the author of the post)

Just as with "you're absolutely right" and the em-dash thing, the image models have a load of habits they fall back on. Some of them are easy to spot - the yellow filter, making grids of icons, ...

Others are things you recognise subconsciously without being able to describe consciously. I've been experimenting a lot, and I can probably see more tells than most people. But sometimes it's just a sixth sense.

And some of those tells do now give me the ick. And some of those tells are in the "better" posters on that page. Today I would clean them off.

But I still think it's unnecessary and dishonest to try and hide that it's made by AI - just tidy up the bits that give people ick. Same goes for text, btw.

Love this. This is exactly the dilemma I had with fiction I had created with AI. Am my I the author? Did I just commission it? Something in between?

So now I don't say. I show my workings and the reader can decide. And to be poncey, I made a manifesto. https://kladde26.org/

When there really aren't any jobs for the former window repair man to do, we are in a "post scarcity economy", which could be a utopia where everyone lives in comfort and has more leisure time.

Or of course, a few billionaires could hoard the wealth.

There's a Wikipedia page about it.

We could be in a post scarcity economy now, except that generally instead of realising we all have enough of we share, we raise the threshold of what we consider to be enough comfort and stuff.

You could, of course, prompt the AI to produce a poster in the style of the "worse" human poster, and see where that gets you.

I suspect it would do quite a convincing job.

Though it's fairly unconvincing at handwriting.

If you break out more sophisticated tools - give Claude Code some skills, let it work in SVG or PDF, create bitmap assets and composite them with real fonts and vector assets, I'm sure you can achieve a lot.

But the point of this blog post was to demonstrate what a PTA mum can achieve even with just ChatGPT, if they just prompt a little more creatively.

The proviso though is that you don't get many tries on free tier. Or if you pay for more tokens, you can fall into a rabbit hole of iterating where it's no longer a quick task.

The sort of use cases I was think of (village fête etc.) would never have paid for a professional designer. We're talking about an event that's delighted to raise £500 for the church roof repair fund. Even £50 for a poster is 10% of the fundraising lost.

If it wasn't AI, it would be Comic Sans and some clipart.

I think you need to remember that most people have no eye for design, and that includes even just presenting the essentials of an event in text.

And AI can actually do it much better than the average idiot. Of they ask.

Like, if you ask - "here's the event details, let's make a poster" in the right way, it's going to come back and say "you've supplied the venue name but not the street address, do you want to include that?".

And then it'll lay the info out with generally sensible relative sizes and positions, better than the average idiot would.

And, probably about equally as well as the average "I'm not a designer but I know a bit" amateur - and a lot faster.

The catch there is that the idiot only gets this if their prompt is right - and since they're an idiot, it probably isn't.

I think there's a gap in the market for a service with system prompts that make it work for people who don't think very hard.

(That might be what, say, Canva Pro does with its AI features. I don't know because I'm not paying for that.)