Now, I’ll be honest. AI art still has its limitations, as my experiments this week have proved. To give a little background, I’ve recently signed up to access a AI art generator through DeviantArt and have been putting it through its paces, testing out the limits of its knowledge base and ability to follow a prompt. The results as you might expect have been a bit random.
In a nutshell, it’s a very sophisticated idiot, to quote Tom Baker. It does exactly what you tell it to. The problem lies with your understanding of the prompt won’t match what it understands. You have to be very precise in what you type and how you separate out the words you use. To give an example, I asked for a picture of two characters hugging. And just by not adding an ‘and’ between the two names, it produced two images of kinda fusion of the two characters.
The takeaway here is that precision is key. And with a precise description it will be perfect for what I have in mind; creating reference and inspiration images for artworks and drawings I am looking to commission. I’m looking forward to trying it out some more.