If designing a new Ducati by asking the computer seems a little far fetched, then think again. In the last 12 months AI-generated art has really taken off. First with DALL-E (named after a fusion of Spanish artist Salvador Dali and Pixar robot WALL-E), then both the Midjourney and Stable Diffusion platforms coming in as entries to get the text-to-creative juices flowing.
The concept is simple: You input a few text commands – called “prompts”, the AI delves into it’s dataset hard drive to search up your shopping list of words then it generates an image based on those same words using stuff that it’s been taught. The better choice of descriptive words and commands, the more likely it is it’ll create something that you wanted.
I’d seen some incredible AI art recently and just had to try it out – I had so many questions. Could it help me with my workflow? Would it replace me as a motorcycle concept visualiser? Was it just a passing fad? Wait… is this image theft?
After a trial run with Midjourney that gave me 25 free image generations I signed up for a month to try it out proper. After a few days watching tutorial videos on YouTube I got stuck in and the stuff it came up with ranged from pictures of my cat wearing medieval armour to Darth Vader riding a Hello Kitty bike and lots of random stuff inbetween. But did I find the answer to those questions? Let’s have a go at answering them, from a motorcycle concept artist perspective.
Would it replace my work? At this stage, the tech is still very new but it IS evolving incredibly quickly. There are tons of glitches in the images it generates – exhausts, handlebars, engines, footpegs, suspension… in fact anything remotely technical, it falls apart. Take the white Ducati e-moto for example. It’s recognisable for what it is, just don’t go examining the detail. Of course a lot of that will be down to my own personal skill level and lack of experience in using the correct prompts. So right now, I think my job is safe, especially if I can unlock best use out of it for my own workflow processes.
I can see this being a handy tool for brainstorming during the early stages of a project, and it’ll definitely be of use during a creative block.
Is it a passing fad? I don’t believe it is, the tech is here, it’s not going away and is only going to improve over the coming months. Is it image theft? When you type in a text prompt the AI has been trained to recognise things like artists’ styles so it can try to replicate that style should you request it. For instance I typed in “in the style of Vincent Van Gogh” for this motorcycle riding on a country road image. Midjourney’s AI has examined images from Van Gogh and replicated his style – all without consent. But is that any different to a random Joe Bloggs with the necessary painting skills doing the same thing?
There are many unhappy artists who would argue that this is wrong on so many levels, because the way the machine learns is so effective that it is not so much image theft but more a case of identity theft. It’s a debate that needs to be had, for sure.
For now at least, I think the best approach is to treat it as yet another tool in the box, and as a creative I suspect I’m in a strong position to get the most from this tool.