Steampunk Fairies

This week, Shutterbug returns to help in some experiments with AI-generated art.

Hello and welcome! I’m Greg R. Fishbone, also known as Cryptoversal, and you’re reading the Mythoversal Cryptoversal Newsletter. You either subscribed or are receiving a forwarded copy from a friend—either way, I’m glad to have you in the Cryptoversal community.

Update: This Issue is now also available as a free NFT on Mirror.

It’s been a busy week of experimentation. One area has been my further tinkering with AI-generative art. The AIs can turn text prompts into stunning visuals, but the best results still require human effort and skill to craft the prompts, guide the process of image selection and upscaling, and edit out the weird artifacts and wonky anatomy that are often included.

Steampunk Fairies Revisited

Wordler fans will remember Shutterbug as the Steampunk Faery from the Samhain 2022 storyline and as a former Cryptoversal mascot. She’s a fun character who is also currently appearing in the ongoing Extra Day saga.

She also loves blueberry muffins.

Shutterbug’s favorite pastime is photography, and her passion is for capturing images.

Previous images of Shutterbug were created to give my stories a more interesting visual presentation, but I was using an older version of MidJourney and didn’t really know what I was doing. And still don’t, in many ways.

In my experiments this week, I tried to hone my skills using some refined prompts and the concept of a Steampunk Fairy as a common baseline.

Anime Style

It seems that MidJourney now has a dedicated anime mode that uses color palettes, body proportions, poses, and styles typically found in anime-style artwork. AI isn’t going to produce animated sequences or consistent character designs across multiple instances, but the results are very cute.

This image could easily be used as concept art to pitch a Steampunk Fairies animated series. Doesn’t this make you want to watch a show about the adventures of Shutterbug-chan and her friends? Animation studios, doesn’t this image make you want to hire me to write some scripts? :D

Etching Style

This style produced an etching-style illustration with intricate details. I could imagine this as an illustration in an old 19th Century Field Guide to Steampunk Fairyland.

Synthwave Style

But if you do visit Steampunk Fairyland, you’ll find your great-grandfather’s Field Guide hopelessly out of date. These days, all the cool Synthwave Fairies will be hanging out at the disco.

Steampunk Neverland

I may have to write this as a book.

The good thing about Steampunk Neverland would be fidelity to the technological aesthetic of the source material’s era. Tinker Bell, as a tinker, would be a natural fit in a world of gears and springs. But poor Captain Hook—surrounded by ticking clocks, he’d go absolutely insane.

We can get a sense from this image how the AI uses images from its training set. Steampunk Tink recalls the post and outfit of Disney animated version of Tink, which must have been included in the training set. Also notable is that Steampunk Tink has no wings, since the AI doesn’t seem to have fully understood how necessary wings are as an element of fairy anatomy.

AI as a tool, or as a collaborator, remains something of a black box. It’s not always possible to understand what it thinks or how it works. But every experiment provides further insight into how to get the best work from it.

When I diffused this image originally, there were odd cables and some weird object was dangling from Steampunk Tink’s mouth. I imported the MidJourney image into Stable Diffusion with the trouble spots masked out, and a further iteration filled in the gaps. The new image acquired a pair of empty eyes, but I was able to cut and paste the eyes from the previous Version 1 into the less-problematic Version 2 to create this composite Version 3.

For me, the ability to edit the images is a real game-changer. It will allow me to make defective images more useful and edge the work closer to the version I have in my imagination.

Are you using AI to generate artwork for your projects? I’m curious what your experience has been.

Upcoming

• As the new Books Editor for Vagobond Magazine, I’ll be promoting and spotlighting the many cool things that Web3 authors are doing.

• The Wordler Village project is progressing toward a next phase that will make it more open and collaborative than ever.

• The Web3 Writers Hour on Twitter Spaces will be happening on Tuesdays.

• The Year in Books project has an end-of-month snapshot approaching. Get yours while you can.

• Cool stuff is being made on Tezos.

--Greg R. Fishbone, the Cryptoversal Mythoversalist