What we learned using AI to design a magazine cover

When create, the multimedia publishing platform we run on behalf of Engineers Australia, explored what engineers need to know about generative AI, it seemed only fitting that we use the technology to design the print magazine cover that accompanied the piece.

How we did it

Usually, pulling together a create magazine cover might mean organising a photo shoot or commissioning a custom illustration.

Instead, we started with Midjourney, one of many AI models that can take text inputs and and produce rich images. We had a clear idea of what we thought we wanted, distilled into a short prompt.

Almost instantly, we learned that with a tool like Midjourney, this approach is more likely to cause frustration than good results.

Images often depart wildly from expectations and instructions. Instead of treating the AI as a faithful robot servant doing your bidding, it’s best to think of it as a harebrained collaborator with raw talent but no discipline.

Instead of trying to bend its output to our will, we used it as an incredibly speedy mock-up generator, allowing us to brainstorm with live images.

Our design and editorial teams collaborated to create literally hundreds of different images. Often an image would be totally wrong but highlighted a path we thought worth exploring further. Many of these paths were dead ends, but some branched into new possibilities, which we could refine and tweak until we were happy.

In the end, the prompt that delivered us the image we wanted came in at almost 150 words, with almost 80 individual prompts carefully weighted against each other.

Here’s a highly abridged snapshot of how it came together.

Once we had an image we thought would work for the cover – a conceivable vision of what an Australian city in the coming decades might look like, as designed by AI – there was still plenty of work in between it and a starring role on the cover.

We deployed a separate AI to upscale the digital image into something large enough to print. And then we handed it all over to a (human) illustrator and retoucher to build in parts of the frame that were missing, to remove strange digital artefacts, and to generally elevate the image to where we needed it to be.

What we learned

Anxiety over what impact AI image generation technology will have on creative designers and visual artists is understandable.

There is also an extremely important conversation that needs to be had around the ethics of training these models on the work of artists without their consent. We are also in resounding need of a set of practices around how AI use is disclosed, artists are credited and how copyright keeps up.

But for now, our view of these tools is exactly that – they are tools.

If the average creative agency is a woodwork shop, then AI image-generation tools aren’t a never-sleeping robot carpenter here to replace the staff – they’re a raft of new power tools.

The right tools – perhaps a table saw, a router, an orbital sander – can make a woodworker vastly more productive. They can dramatically cut down on the tedium of grunt work, freeing up more time for higher-value labour. They can lower the barrier of entry, making tasks more approachable. And they can enable new approaches, spurring creativity and innovation.

The video below is a perfect demonstration of this, and a fascinating look into how an accomplished artist uses AI tools to fuel his creativity.

“I see the overall process as a joint effort with the AI… I feel this is an opportunity, an opportunity for many new talented people to jump on a new branch of art that is completely different from the one that we have already in digital art and just open up new ways of being creative.”

Tractor or washing machine?

It’s easy to see the tale of innovation and automation as one of displacement of labour.

Take agriculture. For most of civilised human history, the majority of the population has worked to bring food from the land to the table. This remained true until the first half of the 1700s, when new technologies and approaches began reducing how many people were needed to grow enough food for a given population. A century ago, almost one in three Australians were farmers. Today it is barely one in 50.

But at the other end of the spectrum is the impact that labour-saving devices – like washing machines, vacuum cleaners and microwaves – have had on domestic labour.

A century ago, the average household spent 60 hours a week on domestic chores. Fifty years later, with a panoply of new household appliances, the average was up to 70 hours.

Why? Because as the use of these labour-saving devices grew, so did expectations around household cleanliness and meal variety. Automation actually increased workloads.

The impact that AI has on creative industries will almost certainly be somewhere in between these two extremes.

People are already using AI tools to mass-produce perfectly average content – literally average, in that the predictions made by today’s generative AI are averages of the data it has been trained on.

It’s the sort of content that pleases search engines more than people, but these tools will likely drastically increase the torrents of new digital content already being published every minute.

Of course, not all content is created equal. As deluges of AI-created content start filling our digital spaces, what will remain valuable and sought-after is truly original content – content that offers something new, something fresh, something surprising, something insightful.

The emerging AI tools make it easier than ever before to create content, but with low effort comes low reward.

This is because the actual production of content – the typing of letters onto a page, or drafting an image into reality – is just the middle step. Just as important are the steps on either side.

First, there are ideas – knowing what content to create and what will engage your audiences.

On the other end is expertise – knowing what separates average from excellent, and how to take content from one to the other.

As the abandoned drafts for our create cover show, if we had relied on Midjourney’s suggestions and judgement alone, our cover would have been underwhelming indeed.

But it also wouldn’t have been possible without AI. And this is what excites us so much about this technology. By augmenting the creative process, rather than automating it, we can create even richer content.