Design

WOMBO Dream AI Art

I downloaded this free app (iOS and Android) that lets you turn an English phrase into a 960x1440px piece of fantasy artwork. It basically runs computer vision algorithms in reverse to hallucinate or dream up new creations. I have spent hours playing with it. It really rewards experimentation with phrasing and styles.

Some fun queries I have crafted:

“Fiery gates of hell vanishing point” “Flaming dragon pedestal” “Glowing God of light sky explosion winged statue”

(Enjoy it while it is free; they are probably burning through VC money and trying to get acquired, or will have to ask for money at some point. That happened to the Prisma AI app that would make artwork out of photos.)

I have played with it enough to get a feel for how to control the results by adding and removing words, and which styles work for which types of subjects. Some useful words to control the general shape of the thing it is drawing: “statue”, “figure”, “dancer”, “orb”, “bust”, “pedestal”. Cool queries to inject elements into a painting: “feathered,” “raven,” “bat-wings,” “devilish,” “winged,” “sky,” “cloudy,” “night time,” “galaxy,” “arabesque,” “mosaic,” “coral reef,” “tapestry,” etc.

I think the strength of the app is the flip-side of its main weakness: it is very abstract and almost like a stroke or a fever nightmare. The strength is that it can create really amazing and surprising results with very imprecise or abstract words (“happenstance”, “dreadful”, “happy”, “funny”, coined words, “reassuring”, moody nonsense phrases like “beyond the tolling bell of avarice”) and particularly, it can combine concepts in a magical way, like “fairy flower,” “winged dancing mermaid,” etc. It can convey a very clear mood or gesture, or set an environment or a feeling.

I like that it is its own art form to craft a query and select and even crop the final artwork. I also like that it is totally random and unreproducible and requires luck to get a good composition. In these senses it is just another tool that I can use to make my own artwork, instead of just producing perfect results that I don’t own, the way Google Image Search might do. The surprise is part of the fun, and gives you ownership of a good “find.”

Additionally, another big strength is its speed and the ability to create many fully-realized, high-level concepts in rapid sequence. Imagine a working artist who gets over the initial response of “I can do better than that” and instead says, “Amazing, a new tool in my tool belt!” That artist could then present clients with dozens of hand-selected small thumbnail mock-ups, and a client could help iterate and rapidly come up with lots of awesome design ideas, on the cheap, in the course of a short meeting or Zoom call, in many styles and compositions, before the artist then goes and paints some real “final” artwork versions, at larger scale and greater cost (of time and money).

It makes me really want to create a card game or board game akin to Magic the Gathering that needs a large quantity of (small) fantasy and sci-fi artwork because it is fun to try to create sets of artwork based on lots of specific ideas, like making a working set of dozens of distinguishable icons that have mnemonic meaning even if they are not concrete or perfect (human figures are a glaring weakness). In fact, the artwork could spur lots of ideas in the gameplay and content side of things.

Archive