Cover Art Tool: MidJourney

I had first heard about this particular AI art tool a few weeks ago, but I’ve been aware of the useful things AI can do for an artist for some time. I’ve dabbled in a few programs, looked closely at one in particular (which needs a much more powerful graphics card than even my gaming-level computer has onboard. No budget right now, but this would not be the first nor even the second time I upgraded computers in order to run an art program), and even with this one, I’d jumped into the Discord community and jumped right back out, overwhelmed. I knew, though, that I needed to get on that wave. What I didn’t realize was just how big this monster surf was going to get.

/imagine: Monster Surf, realistic

Imagine you could type a text prompt into a Discord server, and get back the above image in a matter of moments. The first time I did this, after learning I could use the MidJourney bot in a private server and didn’t have to deal with the madness that is their community server, I made noises no human should emit, and spent the rest of the day in serious existential crisis. As an artist, this thing is a game changer. I spent some time feeling like a buggy-whip maker when Ford rolled the first cars past my shop. As a cover designer, this is immediately apparent as a powerful tool in my hands. I simply could not ignore it. It will be used by everyone and it will change the art world in a matter of weeks, arguably this has already happened. I can’t not use it as a cover designer, everyone else will be. As a writer, the stories…. oh, the stories!

Today, though, I wanted to explore what this can do for you as a cover artist and designer.

Cover mockup using a MidJourney image, a fractal element, and some creativity to combine it all into a cohesive whole that is high-resolution enough to use.

First, let me tell you what you cannot do with the images you generate from MidJourney. You can’t just slap them on a book cover and call it done. For an ebook, maybe. You might be able to get away with it for web resolution. For a print cover, no. MidJourney generates through Discord, and the image limitations will only allow a max upscale of 1664 px square (there are ways to constrain the image output to other shapes, still, they are too low-res for a full cover). So in order to use a MidJourney image, you will need to combine it as an element with other images.

Here I’ve combined two images, with minimal post-processing, into a cover mockup

The other thing you’ll find MidJourney struggles with are faces and human figures. But… I did the above images. Yes, it can be done, just it takes a little effort to get them, and even when you do, there’s often subtly wrong features. Don’t look too closely at their eyes. MidJourney also does not grok animals, leading to some very surreal results. In time, this is likely to improve, as the machine learning behind the scenes is feeding off the nearly a million users of the AI to date.

Here I used two images, one a face, the other a branch of cherry blossoms. The face was flipped, duplicated, and painted over. On the original image the face, while symmetrical, wasn’t usable except possibly for a horror image. Here, it could be a lovely retelling of a classic fairytale as the title indicates.

Where the AI excels is in generating backgrounds. It’s learning from all the myriad images available on the internet, but it is not pulling from those images. It’s using the styles, and this is important when it comes to copyright. It means that the images you generate through use of the MidJourney tool are yours to use as you please, even in commercial applications.

Postapocalyptic Kinkade-style art? Sure, with some additions and postwork to ensure the standard ebook cover size can be met without pixelization!

This is a game-changer for the Indie Author trying to create covers on a budget. Short stories? Buying covers for those makes no sense, you’ll never make your money back. A little work with MidJourney, font selections, and knowing your genre (sorry, you still have to do your homework when it comes to cover art and design! No getting out of that!) and you can have an absolutely amazing cover for very little investment.

The result of asking for a John Singer Sargent style spaceship, plus some painting and a fractal starfield on my part.

As an artist, coming back to that moment of angst when I saw the first MidJourney prompt generate a jaw-dropping gorgeous oil painting style of work that I couldn’t possibly create with my two hands and some tubes of paint… time and playing with the thing for more hours than I care to admit this last week has assuaged much of that feeling of total redundancy. Midjourney isn’t even close to replacing the human element in art. It’s a fantastic tool, and a wonderful game.

“Beacon of Despair”

Ways to use it? Beyond book covers, I intend to use it for blog header art (like, a lot of the time). I’ll be using it for story ideas. I plan to use it as a weekly prompt over at MOTE. For that matter I can use it in my own art as a prompt. Color palettes, composition, design ideas… the possibilities are endless. A friend and costume designer was telling me she follows another costumer who is using it to generate costumes that you could never build with real materials but the inspiration from it is simply amazing. Another artist I know is using lines of poems to generate his art, and splicing them together in sort of visual storylines.

/imagine: steampunk empress, tall, regal, symmetrical, imperious, cloak falling to feet, infernal machines in background, mucha style, insanely detailed, 4K, Octane

Follow me on MidJourney if you start playing with it, and I’ll follow you back. There is so much potential here. The first 25 images are free. After that it’s $10 monthly for roughly 200, or $30 monthly for Unlimited image generations. Very reasonable if you, like me, buy stock art on the regular. Plus, it’s so much fun to play with this! I highly recommend taking time to study the community feed, looking at the prompts, for ideas on how to format your prompts for best results.

Same prompt as the above image, just a new variant on it.
Generated for my son, as he’d just bought his first welding helmet for class. /imagine: welding helmet, rococo, baroque, insanely detailed, welding, frazetta

53 thoughts on “Cover Art Tool: MidJourney

  1. All I can say is WOW! and I’ll be heading over there after dealing with real life stuff today. Thanks for the head’s up.

      1. It is too addictive. I could play with it all day. Right now, I have a couple of covers where I’m not quite happy with the backgrounds and am trying to see if I can do better using the app. Fingers crossed.

        1. I’m very blameable 😀 Nice work!

          Oh, by the way, you can delete images off their server once you make them, which I’m doing with anything I plan to use as a cover or other element. Probably won’t prevent there being a problem, but I prefer to keep my stuff where it can’t be scraped by bots easily.

          1. Yeah, and I’ll be doing the same thing. You are very much to blame for this fun–thanks for sharing it because I hadn’t heard about it.

  2. On GPUs, I would strongly recommend waiting for RDNA3 to hit shelves. Don’t jump on Lovelace. I expect between it and the absolute glut of mining cards hitting the market, you should be able to get even a 3090 for $600-800.

    And the only reason I think the 3090 would stay above $600 is because the 24Gb of vram is useful professional work.

    I’m also hearing rumors that Lovelace Mobile has the potential to be extremely power efficient. While the top end Lovelace cards are going to be cranked up to the max, it sounds like if dial it back, they can actually get really good efficiency, just only at the low end. Monolithic still has significant advantages in low power applications, even if chiplettes are looking much better for massive performance.

    Addendum: I’m expecting RDNA3 to be the better gaming GPU than the 4090, but as I understand it, nVidia has the far better software support in the professional space.

    1. Good to know! With the move still having repercussions on the family coffers (not to mention those welding classes!) it will be some time before I’m ready to build.

  3. I think maybe for someone who wants to draw again, Midjourney could be a drawing prompt? Print out something from Midjourney and use it as a basis to draw from?

  4. Wow! Some of those I’d love to have on my wall. I can see where using this for story prompts or as a visualization aid would be great. I was thumbing through a book of art for the Netflix animated version of *Castlevania*, and found some great matte art and incidental images for “OK, what would the interior of a spooky-but-not-abandoned castle look like” or a haunted town, and so on. This program takes that idea and jumps it way up.

  5. Want.
    So cool.
    Must stick with current learning goals, no changing obsessions until a minimum level of mastery!

    (I mean that in the nicest way, and if I ever get things off the ground, expect that I’ll likely commission some work. Those are awesome!)

  6. WOW! This thing is amazing! Thanks for sharing about it! I’ve tried it out for my Escaping Aurora story cover and may actually have something usable, at least for Amazon.

    PS – I can’t believe you actually did the ‘Beacon of Despair’! LOL

    1. Why not? LOL I am learning and it was grist for a prompt, to see how I could get what I had on my head onto the screen. I really loved what it put out.

      Looking forward to seeing your cover.

  7. Love the work. I really don’t understand the “adding layers, combining images” bit, but then I am NOT an artist, it’s why I find people that can do that.

    1. Well, I know there was a point where I couldn’t afford to find anyone else. I did barter for what I couldn’t do… then again, had I not been forced to learn cover art I might not have gone on to pursue being an artist, which makes me very happy now, a decade later.

      I guess what I’m saying is that l love doing this stuff, but I know not everyone does. And you know I’m happy to work with you any time you need a cover.

    2. “Adding layers” is acting as though your image elements (such as a face) are on a transparent gel, like they would be in animation. That means you can add another layer (such as a background) without changing the initial part. You can also add a filtering layer (I don’t like the color; let me shift it on a separate layer so I don’t destroy the original image.)

      A typical photoshopped art image may have dozens of layers, each doing a particular thing. If I’m working from a photo, the first layer is nothing more than the original photo, and then each successive layer is painted on top of that. You can combine layers if you don’t need to have them separate (saves space, and it’s particularly useful if you’re drawing individual lines that get added down once they’re cleaned up.)

    3. For instance, when I build up a cover based on a public domain illustration, the title and byline are separate layers from the illustration. Then, while I may do some stuff to the illustration itself (remove a color cast and saturate colors), then I put a layer on top and put, say, blue for the sky that is beige in the illustration. This means that I can slop blue all over the sky, carefully trim it to the sky space, and then set to “soft light” to preserve the texture behind. And if I don’t like it, I can do all sorts of mucking about without worrying about the original image being changed.

  8. There are a number of different AI creation places out there, with various levels of user difficulty. If you just want to dip a toe in to see if you would like to try other things. I suggest in order of (my perceived) difficulty:
    Wombo dot art (online and one mobile) – makes “trading cards” but you can “turn off’ the frame
    NightCafe dot com – various sizes and various styles, uses “credits” to make images, but you get a bunch free and they are easy to earn more, I’ve never paid for any credits.
    ArtFlow dot ai – does portraits and landscapes pretty simplistic, but nice images
    Gaugan dot org/gaugan2/ – can do images, text and so on, has pretty good instructions to help.
    ArtBreeder dot com – Doesn’t use text. You pick a picture or elements of a picture and “breed” it with another picture. You can then create new “child” images. Does background, abstract, people, etc.
    MidJourney – They have simplified the user interface, in that you just type in the prompt, it doesn’t have icons for styles, etc.
    DiscoDiffusion – also has a discord, as well as a google collab site. If you follow a YouTube video it’s simple enough, but can get confusing very fast if you aren’t a programmer. But, does some really nice images.
    For 3d world generation: http://www.nvidia dot com/en-us/omniverse/ – Have to have nVidia card, and it’s pretty complicated.

    For one shot elements:
    Thispersondoesnotexist dot com – for headshots of people, also, dogs, cats and horses. – Create planets by changing a few parameters – Another planet generator
    portraitai dot com – mobile only. Head shots again
    There is also nVidia Canvas if you have an RTX card, you make a few sketch lines and it creates landscapes from them.
    There used to be one that could take a portrait and convert it to “historical” styles. But, can’t find it anymore.

    Prosepainter dot com – Upload an image, then type in a line of prose/prompt, then you can paint “it” on the image you uploaded. Makes great magic effects and sci-fi/superhero effects. But, flowers and other things, too.

    For $10 you can get FlowScape which allows you to create full scenes of any size (but not really sci-fi very easily – but possible), You can import other 3d elements into it. And you can then paint over the resulting image.

    There are also free online flame and nebula painters for making backgrounds and overlay elements.

    Hope that helps some folks who aren’t sure they want to get into the “big programs” right off.

    1. Do you know what the licensing issues are for these created images? I played a bit with Wombo, and there were some elements that maybe I would want to use, but I couldn’t find what the restrictions were.

      1. The TOS for MidJourney state the images you generate are yours to use as you wish, to include commercial use. This program is not using bits of other images, it’s creating in styles. The one thing I’d caution about is using celebrity images, which it will generate.

  9. BTW – If I click your “follow” link, it just takes me to my feed. Any idea how to actually follow someone?

      1. I’m running into the same problem with your link. It appears to be the webpage, and not the discord server. I just sent a friend request on discord, and found my way to the main MidJourney page. A bbit of a madhouse, but I’m figuring things out. I’m RHailey KO4LNC#5157 BTW

  10. I know it’s possible to print oil on canvas; we got a small reproduction of a piece the spouse likes. I wonder if it is possible to upscale these and turn them into printable art too?

    I know the cover artists my oldan works with aren’t just selling the rights to use their artwork on the book’s cover. They’re also usually selling prints, digital copies, computer wallpapers, and the originals on their store.

    I wonder if seeding out free, but signed desktop wallpapers would also be a viable way to advertise a print to order art business? I.e. you put up free wallpapers on sites, and make sure the wallpaper itself has some sort of discreet watermark that can be followed back to your storefront?

    Like this scene? Go to our store and you can order one printed direct to canvas and shipped to your door, framed or you can have someone else frame it.

  11. I am curious as to how the subscription service counts number of images per month. Is it all the ones you choose to upscale, or just the ones you download?

          1. Okay, I’m there now. Spent most of my GPU minutes on space battles and Austen fan art, and the rest I wasted. 🙂

            For the fan art, I got some of my handsomest results with “mister darcy regency fantasy franco volpi attore cleanshaven” which is funny because Volpi is at best the Millennium Falcon of on-screen Darcys (may not look like much, but has it where it counts, kid.)

            “alan rickman david morrissey colonel brandon regency fantasy” got me some extremely dour Brandons in weird uniforms.

            If trying to extrapolate the Darcy from the lost 1952 BBC adaptation, try “mister darcy regency fantasy, young peter cushing” – the best result without “young” looked like a middle-aged Darcy who has *all* his in-laws visiting and his teenaged daughters turning up with strange puncture marks on their throats…

            1. *snicker snicker snicker* I’m trying to imagine a young Peter Cushing. And I visualize exactly what you are describing!

  12. There is a PhotoShop trick to try to bring an image up to printable quality output. Blow the image up to full-size at 600 dpi and do a light blur. Reduce the dpi to 300. Convert to lab color. Do a 1.5 pixel Unsharp Mask on the lightness channel only. Convert to CMYK. Done.

  13. Got some really funny faces and a picture with roses and castles where they are the same size. I think I don’t have the knack.

Comments are closed.

Up ↑

%d bloggers like this: