#noslalgia

[therobotmonster]

Making a Monster: Smackoderm (or maybe Ogrephant?)

Technically, this is a WIP despite the main image being complete, because I’m planning on toy package graphic design elements, but this might be an interesting look into my “completed piece” AI process, for those interested.

I wanted to see if I could use Midjourney V4 to produce something resembling late 70s, early 80s toy package paintings, the kind you’d see on lines like MOTU and late-run Micronauts.

Long post after the fold.

To this end, I started prompting with:

bio-mammoth, cyborg mammoth/mastodon/elephant with glowing green bio-goo cannons on its back, huge metal claws, and glowing green eyes. barbarian warrior with green skin rides the beast into battle. battle painting, Masters of the Universe package painting, 1985, by frank frazetta and earl norem, pulp fantasy painting

I didn’t really get my Bio-Mammoth. Complex ideas are difficult for text-to-image at this stage, and I intentionally introduced a lot of conceptual chaos with my image prompts, which included such things as the MOTU slime pit, a beast of the apocalypse, and yes, a Frank Frazetta of a mammoth.

A couple of attempts got me at least a few that were pachyderm-like (the other weirdos aren’t necessarily wasted, as I can adjust prompt and evolve them into their own creations)

I wound up developing about five strains of the concept, tweaking the prompt as I went, until one started to produce results that fired my imagination.

Not just in that it was a cool beast, but in that it had the right look for an early 80s 5.5″ fantasy toy release. A battle-steed creature, with the not-included figure riding on the back (probably in an awkward bucket-seat that leaves little room for organs), four easy-to-lose tusks. Just scary enough to be cool, just goofy enough to be fun.

In all, I enlarged about eight of couple of dozen variants of the image I worked up, and began compositing and repainting. I knew I was going to need more real estate for packaging so there be space for logos, text blurbs, feature call-outs, etc.  So once I got the basic chunks’o’phant squared away, I reposition, and start expanding out and replacing background features.

Then I run into the problem that the humans generated in-picture are all pretty much mush, which isn’t shocking given they’re small portions of the overall composition.

So after upscaling my main composition, I start working on the hyena warrior.

an orange anthropomorphic hyena warrior, gnoll, with a bioluminescent blue spots, shouting, 2/3 portrait composition,, Masters of the Universe package painting, 1985, by frank frazetta and earl norem, pulp fantasy painting

I didn’t spend as much time iterating the hyena, since he was going to be small and I’d have to retouch his hands extensively regardless. Eagle eyed viewers may note that most of what I’ve got here isn’t quite what I requested. AI art is bad as specificity, one of the many things humans do better.

For funsies, here’s a hyena I didn’t use.

I was playing this one free-and-loose, and rolling with elements that looked cool, rather than those that matched my initial vision. If I’d been more intent on a specific result, I’d have prompted more generically and leaned more on manual reworking. This time, however, I was steering with, rather than against, the current, so as to learn how the AI responds to different kinds of requests.

Once the hyena-rider was in place, I did work compositing him in, and did some more background touch-up. After that is color-adjusting and balancing (I like to shift the darker values of the piece toward the blue to add depth), another round of cleanups and tweaks, and then pushing it out for all to see despite not really being sure its “done.”