A few months ago I presented an initial foray into AI Gospel art. Since then the technology has developed even more; still, I don’t think we’re quite to the point where manual-only artists will be completely out of work, but we are certainly getting there.
As far as I can tell, Midjourney appears to be the best publicly available text-to-image program. However, unlike some of the others it’s a little complex to get started, and they only allow a certain number of generations before they start charging money. Still, I thought I’d give it a try with Church-related themes.
Writing the correct Midjourney prompt is an art in itself, and it’s clear that people with formal artistic training are at an advantage here. The way Midjourney is setup during the freeware stage makes you see other people’s prompts and creations while yours are generating, and some of the prompts are quite detailed and sophisticated, so it is likely that a more experiences Midjourney artist could get better results than I did here, but I think some of the failings I’ll point out hold true regardless of your skill level.
To get less serious for a moment, one of the prominent themes in Midjourney creations are fantasy creatures. In Mormon folklore we don’t have a lot of monsters, but I thought I’d give it a shot with the Bear Lake Monster and early accounts of Cain visiting early Church members.
An early newspaper account of the Bear Lake Monster stated that it was “a large undulating body, with about 30 feet of exposed surface, of a light cream color, moving swiftly through the water, at a distance of three miles from the point of observation.” From my myriad attempts it’s clear that Midjourney doesn’t take details about perspective and context into account very well. This shot was probably the best I got of the “Bear Lake Monster” in clear water with rolling hills in the background while being watched by people in 19th-century garb.
I tried multiple times to get an image of David Patten’s visit by Cain (made famous by its reprinting in The Miracle of Forgiveness), but getting a man on horseback with another man standing next to him, and describing the two men in detail, was too complex for Midjourney to handle and I couldn’t get anything usable. Instead I created an image of a Cain visit based on E. Wesley Smith’s purported account while he was a mission president in Hawaii (which is anonymous and based on who-knows what, but is still fun).
Of course, here the PG-13 filter gets in the way since the account has him naked. Also, given the time period there was probably some racist assumptions vis-a-vis Cain, but I’m going to take advantage of the fact that this account doesn’t specifically mention Cain’s race.
On a more serious note, attempts at visualizing absolutes that are supposed to be indescribable such as hell and heaven have always been fascinating to me, so I tried some basic prompts, and they turned out okay.
Two renditions of “Outer darkness” and “Hell,” according to Midjourney.
Heavenly beings on a sea of fire and glass. (Still a little too dark, I tried to brighten it several times but couldn’t quite get it).
“Descendants as the stars of the heaven and as the sand which is on the seashore.”
Finally, the point has been made elsewhere that, as the Church diversifies, Church art could become a little more diverse, and AI helps people create variations on similar religious themes. I asked Midjourney to create a “Japanese Eve eating the fruit in the style of traditional Japanese art.”
As of now AI generated art doesn’t do text well, so somebody more informed than I will have to determine whether the Japanese characters are gibberish. Also, perhaps because of the PG-13 filter, the other options included scantily clad versions of Eve that were in a way more erotic and inappropriate than had she been portrayed in the more traditional bare-breasted or foliage-covered way. (Not that I’m against the filter; without it we are really at the point technologically where some creep could upload a photo of the boy or girl next door and impose them on some sexual fantasy. On a related note, in between the first and second draft of this post a Midjourney-type service for movies was announced.)
Presumably you could do other similar variations: “Eve as an East African woman eating the fruit,” “Eve as a contemporary Peruvian woman eating the fruit,” etc. If the point is to stretch our artistic and conceptual paradigms beyond Greg Olsen, the variations on different religious themes have suddenly become endless. Additionally, more advanced prompt artists can upload other images and generate variations on a theme, and this is where I think manual artists may be able to use AI as a complement, instead of a substitute, for their own work.
In summary, AI is getting close to human-level for basic stills and tableaus. However, for complex context or images outside of the standard range (I tried many times to get a city of LDS temples and it never got even close) there’s still some work to be done. However, at the rate things are going I don’t think it will be too long before those gaps are closed. Whereas in the past we were limited to the relatively slow production of galleries and art shows, in the near future I think skilled religious prompt artists are in a position to literally create millions of pieces moving, inspiring religious art across the whole range of styles and preferences.