The new ChatGPT Images 2 is now available across all the different ChatGPT subscriptions, including Free. It’s a much improved version of the original ChatGPT images that features more precise image generation, particularly of small text, more aspect ratios for images, and paid users now get thinking-enabled workflows, so it can reason through an image generation task, and even search the web where it needs to find out more background information.
One of the things ChatGPT can now do, which really grabbed my attention when Sam Altman demonstrated the feature in his introduction video, was spin up what look like real magazines from text prompts and photos. ChatGPT can make magazine covers, and also whole spreads from the interior, like contents pages and editorials.
I edited print magazines in the 2000s. Among those, my longest stint was on Apple magazine MacFormat, which is still going strong today, so this feature obviously caught my eye. And it also raised an existential threat — are magazine designers going to be another victim of the AI jobs cull? Let's find out...
How to make a magazine in ChatGPT
To make a magazine cover you just give ChatGPT a picture and ask it to turn it into a magazine cover using a prompt like:
The prompt is: “I want to turn this into a magazine cover featuring these two men. One is Tom[sic] Cook (on the right) and John Ternus, is on the left. Tetanus[sic] is taking over for Cook as Apple CEO. So let’s see a dynamic business magazine cover.”
In response, ChatGPT makes something that looks like a well designed magazine — at least on the surface.
Generating text used to be the real Achilles Heel of AI image generation, so this is seriously impressive. Text in AI images used to routinely degenerate into gibberish symbols, but now it’s crystal clear and on point. But then you start to notice the little things…
For example, there are full stops at the end of sentences, which you do see on magazine covers sometimes, but isn’t the usual magazine style. And the cover lines are repetitive, which isn’t surprising given that there’s not a lot of information for ChatGPT to go on. If you were actually going to use this image for any professional project, you’d need to start making changes, and that’s where things fall apart.
The big problem is that this is a graphic, so I can’t just load it up in InDesign and edit the text as I normally would with a magazine cover I was working on. And that got me thinking — what then is the point of this?
I quite often have the same reaction to all the impressive things that AI can do, especially around images — what is this for? Who is this for? Why would you want to do this?
Recently, Google announced that Gemini can now access your Google Photos, so it can reproduce images of you in familiar situations similar to those in your photos, but in a cartoon style or with the people made of plasticine. I mean, it’s very impressive, but why?
I guess you could share the images with your family members, but that’s going to get very tiring, very quickly. You can imagine the rolling of teenage eyes — “Uh Oh, here’s Dad with the AI slop again…”
I have similar thoughts on this magazine cover. Maybe it’s useful as a proof of concept if you were planning on making a magazine… but how many people using ChatGPT are thinking seriously about making a print magazine? And if they were, they’d be making something in proper software, like Adobe InDesign, where you can edit the text in and move the headings around.
Can ChatGPT replace designers?
I wouldn’t be surprised if, after the ChatGPT Images 2 product announcement, there are people going around saying that ChatGPT can now make magazines, or posting the images of the covers they've created and saying things like, “Bye-bye art directors”. But we’re simply not there yet.
Producing an image is a long way from making something that could fit into an actual magazine production workflow. It’s only when you dig deeper that things start to fall apart, which is true of so much of the content produced by AI.
I’m not expecting ChatGPT to make something that a proper DTP program like Adobe InDesign could, but equally, I’m tired of all the headline-grabbing “now AI can do this!” features that turn out to be aimed more at impressing shareholders than providing something practical and useful. And that feels like a pattern that’s starting to define a lot of AI announcements right now.
Perhaps I should be thankful that AI can’t replace me just yet, even if it’s very good at generating things that look like it should be able to.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.