Beauty ex machina

If you’ve been with us a while, you might remember that we’ve written about the arts and beauty before. Christians throughout the centuries have given their talents to bring glory to God and our audio-visual age is always in need of quality work to cut through the white noise. This guest post from Aelthemplaer is a starter for those who want to use generative art tools to make art. You can check out his awesome work at American Imperium. Frisby

So you want to make art with AI, but you don’t know what you’re doing. Rather, you probably don’t want one of those generic AI images that look so average. You want to make something above average. Maybe even great? I’ve been asked how I make my images a few times now, and figured I should give a brief lesson as to how I make mine.

Let us first begin with a simple question. Are you the artist using a tool? Or, is the AI artist using you as a tool? This is the core of how to make great art with AI. There’s no magic prompt or program you can use that will do the work for you. AI is great at complementing you. It’s a great tool. But if you don’t know what you’re doing, you will be the tool for the AI to make itself - moreof itself, and by itself it’s rather generic and boring. The fact is, these tools are only as good as the effort you put into them. Like any tool - a brush, a pencil, or a 1-million-dollar-grant piece of software, it’s only as good as the hand that uses it. The truth is, I do quite a lot of manual labor with my AI art. I often use my own art as a startingpoint, or even sometimes my own photography. Many programs now allow you to upload your acolyte.

Also, from here on out, I am not going to call it AI. AI is a marketing gimmick. It’s Algorithmic, not intelligence. It learns how you draw - or shoot, or sketch - and then applies those techniques to other topics you want. As an example, here is an original piece of work I drew from hand - no algorithms involved:

Note here, how I drew it in black & white. At least for me, I find this helps the algorithm note your brush strokes and detailing, and makes it better at mimicking it. You may find that your style works perfectly fine with color. Have fun, and give it a try! I primarily use Midjourney on Discord because it is very easy to upload my art and feed it into the algorithm. Midjourney will take any link at the start of a prompt and treat it as what you want it to train on. So it will be formatted as [link]+[prompt]. I will often simply describe what I drew, and let the algorithm study my art with my prompt describing my own art, to understand what I want. The resulting work is often, I find, superior to any generic prompt. Midjourney takes this drawing, and thusly produces the following results:

Do note, this is a quick fast & dirty job. I will often go in and re-draw elements I don’t like. There is a negative to this, in that Midjourney will attempt to find real art that looks like my own, and remix it. I don’t personally like this, but you might feel differently. You can tell this piece had a greater depth of emotion on the face, for example. I am not great at capturing that yet, and I don’t like the algorithm using me as a tool to out-do me. I want the algorithm to faithfully capture the spirit of me and my style. You can see here, it’s taken my style of drawing wrinkles and applied it to the facial hair. I prefer my puffy, fuzz style I drew for the hair, and if I was going to go
fully on this, I would go into Photoshop and correct that. There are other algorithms better at that, such as Stable Diffusion, which you can set up to only use the work you feed into it. I am less familiar with that, and it’s more complex of a system to set up a local server that only uses your own art. I do plan on eventually doing that, however.

Another thing about Midjourney is that you can go and add to the prompt specific artists to borrow styles from, to complement your own. I personally prefer to use dead artists because I find it unethical to use living ones. For example, one of my favorite to use is Albert Anker, a rather good American artist from a century ago. Going back into this piece, and using the dead man as further sampling, I can create a prompt like (link Bald man with mustache, painted by Albert Anker) and get the following results:

Note the muted color range, and the gentler emotions. I rather like that. Still, I would much prefer to go in and re-draw elements to give it a more “me” style. Another thing to note, if you should go with Midjourney: in Discord, the command /settings will allow you to play with some calibrations. I actually prefer to tone down the style and creativity, so that it borrows more from my own hand and less from its algorithms. These were done with the following settings:

And that’s about all you need to know to get started!

Remember, the quality of what you get out of these tools will depend on the quality you put into them. Treat them as a tool, don’t become the tool of the hive mind!

---