I haven't played much these last few weeks. Well, that is, if you except generating "art" on Midjourney. The art-making AI has now reached almost a million users on its official Discord server. What was meant to be a tool for graphic designers, artists, and content creators is progressively attracting a more casual audience of curious people like me, just thrilled at the idea of becoming part of this new step in history when algorithms can dream of electric sheep.
This casual audience stays engaged because Midjourney feels so easy and unengaging to use, and catches us in an incredible feeling of power and flow, like the most addictive sandbox games do.
Midjourney has it all: a solid core loop that lets you "imagine" anything, clever nudges that work like charms, and a community-based, open culture that makes it easy to iterate, learn, discover and progress.
Instant fun, instant reward
So, after having seen some experiments, notably by French designer Etienne Mineur, I jumped on the generative art wagon a few weeks ago. I started off with Disco Diffusion, and was fascinated by the way it could generate landscape images. Then I got the invitation to Midjourney's Discord Server (not yet public at the time), and it didn't take long for me to be caught in. First I treated the same prompts in Midjourney and Disco Diffusion, with the clear-cut idea that I wanted to make some science (or something like that) and get a fair comparison of the two.
But that didn't happen. I just hadn't understood this wasn't the way Midjourney is meant to be played. While the opaque environment of a Google Colab clearly makes Disco Diffusion a connoisseur tool, and can even feel intimidating, Midjourney lets you forget the whole complexity of its dreaming machine.
When you arrive on the Midjourney Discord server… Well, let's stop there. It is, indeed, a Discord server. Not an interface with loads of parameters to change and experiment with, not even a clean UI with some easy shortcut buttons (like Dall-E mini or Nightcafe). It's not presented as an application but as a conversation. At the end of the day, using Midjourney is just the equivalent of asking a friend "what do you think of...". A more "tool-based" app would probably make us more careful about the way we spend our first free credits (trust me, it'll only take minutes before you run out).
As opposed to the long hours spent trying to understand how Disco Diffusion works, the sheer simplicity of Midjourney makes you immediately lose sight of what you went here for (that is if you had any exact plans in the first place). There is no need for a tutorial, just type along. Generally, the first thing that comes to mind, is no work and all play.
There are in fact very few bad results on Midjourney. You can like one of the four images presented to you and draw on one or more of them, and most of the time there will be one nice result that encompasses at least part of your prompt's elements and will feel rewarding. The worst that can happen is something a bit messy or nightmarish that might just make you want to find a better way around your prompt (first incentive to level up!). But let's be honest, Midjourney has such a focus on consistency that it will always prefer giving something that looks like art and avoiding dealing with the hardest part of your prompt.
Midjourney's chore mechanic is a perpetual loot box of sweet, well-rounded art visuals.
Typing random things and testing the limits of Midjourney's resources keeps you focused and in the flow. While my experience with the long hours trying to understand Disco Diffusion parameters had been sometimes great, sometimes frustrating, my first hours, days, and weeks with Midjourney felt increasingly rewarding and fun. Just the two things you'd ask a game to be.
Being part of the chanting crowd: the social dynamic of Midjourney
Be it a beta limitation or not, the fact that Midjourney is hosted on Discord reinforces its playful (almost addictive) nature. We are invited to treat it primarily as a social space that happens to host an incredibly powerful artistic AI rather than the other way round. The choir of ideas being generated at once by all the users is almost overwhelming. But not like this time you were in Athens' Parthenon and went all agoraphobic; more like the times, you were chanting in a festival crowd.
The command bot alone could keep you focused for a long time, and your progression could as well be done by self-documentation or by using forums alone. But what makes Midjourney stand out is the fact that you are a de facto part of a community. You have to accept the fact that your work is shared with everyone, and what bothers you in the first place becomes very soon a source of inspiration and knowledge. It makes it easy to follow the AI's adjustments and new features, easy to contact someone whose prompt you liked, easy to copy-paste some keywords that seem to work well, or even "steal" someone's prompt by directly asking Midjourney to upgrade one of their images.
As a side note, let's mention that the loading times themselves are perfect. This might not be intentional, but while Disco Diffusion will oblige you to leave your computer alone while it computes, Midjourney's content generation only takes a few minutes. Just enough to see the magic progressively take shape, and just enough to have a peep at what others are doing, yet again keeping you engaged.
Being there among the crowd is the main progression mechanic. As a beginner, you just don't yet understand that the simplicity of the prompt can be refined by multiple modifiers that will help you change the image ratio, choose a different style, makesure a portrait is well centered or help you leave out the blur. This is how you progress towards the different levels of Midjourney, which are both levels of knowledge in the "prompt craft", but also the knowledge of what Midjourney is likely to do well or to struggle with.
Frustration and progression
Of course, as in any game, this one comes with some frustration, especially once you understand Midjourney's limitations. Very long prompts generally result in half-baked results. Having two precise individuals doing precisely what you ask them to in a very precise style is probably akin to the final boss. The AI will take some of the elements and exclude most, working towards something that will be visually pleasing and (probably) excluding the elements that don't fit its knowledge.
Thankfully for visual artists of today, there are still some complex ideas that require their talent.
I hope I convinced you that Midjourney should be your next sandbox game. Join us and participate in the perpetual brainstorming. Let's forget for a minute the more complex philosophical, societal, and legal questions this will pose in the very near future... Actually, come and fully engage with them on the very same Discord server, while Midjourney is "imagining" a giant tsunami of broccoli flooding Paris, in the style of William Turner.
Sign in or become a SUPERJUMP member to join the conversation.