You may have seen some sinister 9-square grids f that contain abstract images of ‘Machine Gun Lorraine Kelly’ or ‘Greta Thunberg setting fire to a forest, these aren’t the artworks crafted from the hand of a master photoshopper, but instead, it’s an AI creating generative art from text-based descriptions.
This new trend is being pushed by a couple businesses, but two are leading the way; Open AI’s ‘DALL·E 2’ and Google’s Imagen, both releasing new versions a month apart in spring 2022. Whilst the latest developments by both are unavailable for public use, they have each released examples and research papers to prove their findings.
This powerful technology can be used by brands and creative industries to create more than the meme-worthy images we’re seeing across social media. For now, you can try a version of DALL·E for yourself with this Dalle-Mini, built by Boris Dayma.
The results are mixed with the most successful looking like the inner workings of Francis Bacon or Edvard Munch, but you can see how the latest iterations have started to achieve a realism and accuracy that is mind blowing.
What it is…
Open AI is a research group that has given access to an API that allows you to understand or generate natural language or code. This API paired with a huge data set of images and machine learning algorithms is what makes DALL·E possible.
But that doesn’t mean that text to image is the only use case for the technology, just the one that is hot right now. There's scope for the API to help with searching through the content easily, and classification of content and in the creative space, we don’t need to be restricted to images, but can be applied to generative music and sounds.
As with any creative automation, there's a fear that this technology will replace the care and craft that comes with image creation and storytelling. However, there’s also an argument that AI generative image creation will be a catalyst for creativity.
It not only democratises creativity so that if you have an idea you can bring it to life quickly, but it can also be a building block or foundation for further expression. Imagine you’re a 3D character designer, but you want to place your characters in a beautifully painted world – you can now describe your scene, refine the image and place your characters in the world.
It will revolutionise image searching and world-building, as we step further into creating new realities and the effort required to make these expansive environments, generative images could help create art for 3D spaces quickly, imagine creating a haunted house and needing to create every horror-filled oil painting portrait, now you can generative thousands of assets quickly and easily – in the future, we could also see these assets becoming more personalised to the user for deeper and richer storytelling.
We’re excited by the opportunities this brings and the potential use cases, watch this space for more updates.