Friday 26 January 2024

The Artwork Revisitation

 If you've been reading this blog for at least a couple of years you might remember this bit of acrylic art I put up early in 2022:

It was OK but something about it wasn't quite clicking with me. Then I read a recent article in The Guardian which included the brilliant Chris Foss talking about AI-generated imitations of his own style, and how they could never be mistaken for the real thing.

I looked at the images in the link to Midjourney provided in the article and agreed that they only looked superficially Chris Foss-like - big bristly spaceships floating over alien landscapes etc - but the particular thing that the artist noted was that the images lacked depth, something undoubtedly characteristic of his pictures. He nonetheless noted that the AI-generated images might serve as useful prompts for composition and lighting, so he wasn't dismissing them entirely. 

It occurred to me then that one of the faults with the painting above is the absence of depth - there's some atmospheric misting to push the towers back a bit, but the spaceship is just floating there more or less side-on, with no real reference points to place it in the scene relative to the other elements. Sometimes I use one or two point perspective construction lies to give a sense of a spaceship emerging from a scene, but in this case, I just winged it and drew it without any reference to perspective, figuring it would come out all right. I'm still happy with the ship, but in light of the Foss article - and a bit of renewed mojo for breaking out the airbrush - I thought I'd take another look at the canvas. And, rather than show the finished result, I thought it might be more fun to illustrate the process, even if it all ends up going horribly wrong.

So here's where we are now:

I did a bit of additional masking and spraying on the ship itself, bringing out its forms a bit more clearly by defining shadows and highlights, but the main thing has been to start work on some foreground elements which project out into the scene and serve to push the ship back. In this case I picked a single perspective point and drew some lines projecting out to the vanishing point, which (when finished) will be walkways or landing pads of some kind. I could envisage some smaller spacecraft and/or figures in the nearer foreground.

I also added a touch of contrasting colour in the sky, but this came out much too heavily and will need to be pushed back a bit.

I still don't know whether this will end up being a piece of art that I'm satisfied with, but the journey is fun and these corrective steps and additions can be very educational, so even if this painting fails, the lessons learned will hopefully inform the next one, and the one after.


  1. The creative process is more often about the journey...not the destination. Your update drawing gives the image much more depth. ~Neal

  2. Nice to read that a fellow author is also into painting and drawing. I got a Wacom Intuos for Xmas and when I find time to get to grips with learning how to get the best out of it have plans to draw spaceships etc.

  3. Cool progression, looking good!

  4. Nice work. Second iteration gained a bit of depth so you are definitely heading in the right direction. Nice to know your other talents :)

  5. "Looks like a fish, moves like a fish, steers like a cow".... ;-)

  6. I liked this one when you first posted it and I like the changes. I still think it looks like some sort of sleek submarine in the oceans of Europa or something, maybe a liner with all those big windows.

    As for AI art, I don't buy most of the criticism. I've played around with Dream and Midjourney and used the former to make clip art for my blog, and the progress of its algorithms in one year has been astonishing. I've asked it to make different things: Movie posters for classic 1950s science fiction films, lush alien jungles, comic books drawn in the style of Jim Lee circa mid 90s, photos of eerie suburbia at night a la Gregory Crewdson, even the Spider Room from Revelation Space. It took me dozens of revisions and many more adjustments to the prompts before I finally ended up with the Spider Room that is the product of my imagination interpreting your prose: All brass and red leather cushioned seats, the high tech mixed with the low, rivets and wooden accents, and large windows with views of black interstellar space.

    Most of all, I enjoy making absurd images of my cat. My cat as the King of Joseon circa 1600, dressed in red silk robes with embroidered gold dragons. My cat on the dance floor at a throwback disco party wearing a tight black muscle shirt and surrounded by beautiful women. My cat speeding along atop a futuristic motorcycle in Akira's Neo-Tokyo at night. My cat with a big smile on his face, swimming through a pool of catnip.

    As far as absurd hobbies go, I could do worse I suppose.

    Anyway, the point is that the algorithms have gotten so good that, often, their output is indistinguishable from real artwork, and they will only continue to improve.

    Whether that's a good or bad thing, I can't say. It's not unique. Plenty of voice actors will be out of work with the incredible AI voice synthesis available now. Wedding photography, except for the high end, mostly died out with the advent of cheap DLSR cameras with AI-enabled assist modes. Software like Guitar Rig models guitar amplifiers while VSTs model classic analog synthesizers with such accuracy that they are indistinguishable from the real thing, especially sitting in the mix. You no longer have to dream about getting to use a vintage synth from the late 70s that goes for $50,000 now, you can buy a $99 VST plugin that models it precisely.

    Things are changing and there's nothing we can do about it. The only thing we can change is how we respond to the changes. Thankfully your field is safe for now and the foreseeable future: Chat GPT et al can make passable junk content to fill websites, but they can't produce effective prose (yet) and they can't mimick the imagination, structure, pacing, instincts and narrative skills of a good storyteller.

  7. Rather than generating images via prompts, I suspect where AI will be useful is as assistants in working on parts of art creation - changing something through description - sort of like what Photoshop does now. Blender and Unreal engine getting these types of AI plug ins will lead to interesting results.