Author: DK

After about 20mins of reading and prompt writing, these are the first images I created on the AI-assisted-image-creation-platform Midjourney.

The fidelity of these creations are incredible. Check out these closeups:

As a further experiment and using one of the already created Celtic warriors as the basis for the design, I went on to develop some potential covers for a fiction book I’m writing:

The reason for exploring this new platform is to understand the potential for creating images for my own creative projects and also is this something to build into the public speaking masterclasses I deliver for clients.

In a part of these masterclasses I talk about and show how to craft kick-ass presentation decks. In doing so I utilise a vast array of open sourced and creative common licensed content available on the web via a sweet bevvy of amassed resources over the years, which means that within twenty minutes or so I can find some wonderful items to embellish any story being told and build out a pretty wicked looking deck.

However, after just an hour or so playing on Midjourney, I see it becoming a powerful part of a storytellers toolkit as the business case will be: why trawl through different sites if you can very quickly state your needs and get a high-fidelity version of a specific image which serves the need.

Examples

Whether you’re an architectural student talking about the parametric abilities of your discipline:

A team leader wanting to represent a metaphoric path relating to a bold business strategy needing to be pursued:

A human resources lead presenting to the board and needing a stock-style image of people laughing in a corporate setting around a computer to demonstrate the aim for a conducive work culture:

Or just having fun turning one of your favourite cartoon characters from childhood into a photo-realistic portrait:

The only restriction is your own imagination (which I why like the term Augmented Imagination for AI)!

Limitations

It’s not a perfect platform.

Here’s what I got when I asked for a simple yellow post-it note with the word ‘help’ written on it:

There’s also a barrier to entry from both a digital interface perspective and also an image literacy angle.

Access to Midjourney (which is still positioned as being in beta) is via the Discord platform. This means the creation and indeed craft of producing these images is done out in the open in the semi-public hosted forums. I say semi-public because without first a Discord account plus accepting a few T&C’s to access, you can’t play.

Then there’s the literacy step. Knowing about cameras, lenses, exposures, angles plus lighting will give your images the edge above others. You have to be open to the messy process of nudging the system in the right way and nuances of language along with scene setting or even world building is needed.

Whether it be creating cute little monsters:

An atmospheric and rainy backstreet in an Asian city:

To a toy figure of Michael Jordan:

…you have to guide the output heavily.

The Trick

It can all sound too good to be true, or for some I’ve spoken to before I started playing, a little too confusing as well.

Once you get your head around the interface (as it’s quite jaggedy in it’s forum style, think ‘Slack channel peppered with a cacophony of people you don’t know saying wild and wonderful things with images blooming to life in between your prompts’, as it scrolls without you and finding your creations becomes sometimes hard work), I found the trick in creating impressive images is to balance simplicity and complexity.

Laying down the basics along with the intricate details of how to construct the output you want to see yields high-end results. For example, to build the image above I stipulated the scene, the person, the emotion displayed, the specific laptop being used, then the angle, the camera, the lens, the exposure, the lighting and the general ‘feel’—there’s a lot steering the production of what you see but it’s worth the time.

Built in to the Midjourney service are lots of ‘commands‘ to get your head around, and an interesting one is the “/describe” command which writes four example prompts based on an image you upload. So if you’re already using an image that you own or created you can upload and use the manufactured text prompt to get a sense of how you can create another one like it.

This and many other commands is a paid for service (as part of Midjourney subscription plans) and as I’m still using the free service I point it out as a great way to get started along with finding inspiration in prompt writing approaches.

Ownership

The current T&C’s of Midjourney says you keep what you create:

“…You own all Assets You create with the Services, to the extent possible under current law.”

…although it also states they own it all as well (including the prompts and any reference material you upload) along with some other funky stuff in there, like anyone else in the community can use them and remix them also.

It’s a bit confusing.

Just be aware of the classic adage: if it’s free or cheap you’re probably the product at this stage.

Again, if you want to create in the dark (without being in the public forum) you have to pay for that privilege.

Ethics

Like most though I’m worried deeply about the ethical nature of these creative engines due to questions on where they sourced their data (the large language models, LLMs, used to ‘train’ the AI on—I recommend reading A Completely Non-Technical Explanation of AI and Deep Learning if all this is gobble-dee-gook). Plus, if any output is being influenced by specific artists or other creatives, where is the attribution and citations for them? Surely that should be there as standard and would be easy to code in if you’ve come this far…?

There’s also a question of fabricating ‘alternative’ realities (like the recent Pope in a Prada coat). But what about making different historical narratives?

Like in the latter years of his life, Steve Jobs spent most of his time in his small home office working on old school computers:

Or photos of Elvis when he was a baby already playing a guitar:

Then there’s other questions including environmental, as there’s valid questions around the growing carbon footprint of these platforms plus as mentioned, unethical concerns regarding image scrapers who took 30 billion images from Facebook and other social media sites and gave them to cops: it puts everyone into a ‘perpetual police line-up.

Conclusion

I’m dazzled by this!

It’s captivating to be generating such quality with little effort. And there lies in the danger. To not respect the background nature of the inventive act of creativity is what I fear people will miss here. I’m lucky enough to be able to navigate through my understanding and respect for what informs these constructions (I mean the influences from photographers and my basic knowledge of cameras plus image form as well as the technological interface), and yet it’s so liberating at the ease by which impressiveness can be gained.

With all technologies, I always experience it through the lens of augmentation. Does this aid what I do and offer into the world, with kindness and deep consideration on the impact?

Midjourney and the larger AI suite of tools (currently) do nothing without us. They are value-neutral. However, it’s simple to see how much negative impact they will have in the future. Have us all questioning what we see whilst bad actors look to exploit this to their advantage. And I mean much more than cheeky photographers winning prizes using AI.

So for now I invite you to take some time and look at the public Midjourney showcase to see the possibilities of what people are producing, it’s very hard not to raise your eyebrows in awe.

What about you, what potential do you see in platforms like this? Where will you use it? Why won’t you?

Because you never know when you need to create a shocked looking blue faced pink sheep in a meadow:

All images produced in this blog post was done by Midjourney and the prompts I came up with.