TIME TO READ: 2 MINUTES

I signed up for a Midjourney course that claimed to teach me how to create a “beautiful piece of AI-generated art”. However, I was somewhat sceptical about the process. Could I really create accurate images using only text prompts, or would my creation end up looking more like Frankenstein’s monster than a masterpiece by Frida Kahlo?

Setting up Discord and joining the Midjourney Server are the main prerequisites to start creating AI artwork. To gain access to Midjourney, you need to install Discord and set up an account. Navigating to the Midjourney homepage in Chrome, and signing in, your info is magically beamed to your Discord app.

You then have the chance to give yourself a display name – “Design Voyager” perhaps? Follow the instructions for installation and then “claim” your account. Once verified by email you are ready to create your own server and “invite” the Midjourney bot to play there. You can actually invite others to your server and they can interact with the Midjourney bot themselves. At this point, I realised the basic plan was not going to be enough for what I wanted to do so I was compelled to hit the subscribe button and make a purchase!

Using the /imagine command in the prompt box, I inputted a request for an image suitable for promoting a fictional AI conference. This generated 4 results of varying suitability – see image below. I selected the fourth option (U4) to be Upscaled for better resolution. The upscaling options take the image from the 512 x 512px default up to a maximum of 2048 x 2048 px – which works out at 173.4 x 173.4mm @300dpi so not ideal for large-scale print. Although you can use programs such as Gigapixel AI to enlarge but at an additional cost.

I hit the Remix button to alter and enhance the prompt or type in the prompt bar /settings to access more options. I scaled out by 1.5x or 2x to add to the image – a function very similar to Photoshop’s generative fill, and used the Stylize options to create more variations – the higher the stylized settings the crazier the output! It’s also worth noting it does take a bit of time to process the final results depending on the complexity and current demand.

There is a bit of trial and error involved, but I believe that’s what makes it challenging and appealing. While there are ongoing legal debates regarding copyright, would it be useful for commercial images? Possibly for digital output and definitely valuable for generating ideas, concepts and storyboarding. I’ll have to make use of that subscription and explore more!

©Jennifer Martin 2024