05 July 2022

The Silent : More Experiments in AI

Stuka

I have been experimenting more with the Midjourney AI, trying to get a feel for what I can do with it. I am more of less convinced my use for it will be for environments that I can incorporate into 3D montage work, especially for ‘The Silent’. I am also finding that to get the best image or, more to the point, an image I am happy with, takes post-processing work, often building a composite image and manipulating it to achieve the desired end result.

Unsurprisingly, when it comes to designed objects, it’s a very hit and miss affair, more often miss than hit… like in 95% miss. The Dall-E2 engine seems far better at that, though its outputs are nowhere near as sophisticated as Midjourney’s and is quite limited… in as much as a machine doing design work can be limited!

This image was done on a whim, after a lot of futsing around trying to get Midjourney to generate something else entirely; the algorithm is by no means fool proof and learning when to give up is a key skill one needs to develop. I threw in a prompt more or less as a ‘I wonder’ type experiment and after running through variations and refinements, started to see something that I liked. I then took the five best into Affinity Photo (the Photoshop killer if you’re wondering) and produced this image. There was something about the ‘aircraft’ form that I really liked and after some work, got it to be, at least to me, quite interesting.

While this is quite off the mark for the imagery I am developing for ‘The Silent’ series, what it does prove is that using Midjourney, and/or/maybe Disco Diffusion, to generate aspects of the images I am working on is definitely a valid idea.

0

Copyright 2022 Gerard Thomas. All rights reserved.