Text to art Tanya Combrinck looks at how AI tools are supercharging creativity and taking the digital art world by storm
Over the last few weeks you may have noticed a new kind of artwork creeping into your Twitter feed; fascinating, highly detailed images with composition and style quite unlike anything that’s gone before. It’s Midjourney: a new AI tool that creates art based on a text input. Text-to-image AI tools have been around for several years now, but this one is attracting particular attention for producing exceptionally beautiful and original abstract artwork.
Midjourney is currently in private Beta and runs as a bot on Discord. You can apply for access over at midjourney.com. To start creating images, you join one of the channels and submit text prompts in the chat window. You might try something like “a beautiful painting of a steampunk owl in the moonlight”. The AI bot replies with four thumbnails, and if one or more are close to what you are trying to create you can choose to iterate on them to receive four more thumbnails. You can continue iterating, or you can tweak your text prompts to get different results. When you have an image you like, you can upscale your chosen thumbnails to get a high-resolution output.
Anyone can get decent images out of Midjourney by entering simple descriptive phrases, but to achieve something special you need to be imaginative and learn how to use the prompts to get interesting results. A major plus point of the tool running within Discord is that you can