Imagine opening Google Images and searching for a nice cat picture to accompany your latest article. You would come across countless results, but likely very few (if any at all) that you could legally use due to copyright. This is where a good graphic designer or a stock image site would come in, but usually for a hefty price.
Image generating AIs such as DALL·E 2, Midjourney and Stable Diffusion allow you to type in a ‘prompt’ much like you would in a Google search, to generate a selection of results. However, the distinction is that you are given ownership of every picture that you generate. Stable Diffusion can be used locally for free if you use your own hardware to run the AI, whereas DALL·E 2 and Midjourney are available online for a small fee.
Given that these AIs are substantially cheaper than the traditional means of sourcing images and graphics, you may be wondering if this reduced price is also indicative of their quality. Uniquely, AI image generation stands as an outlier to the usual price-quality relationship. For example, an art piece generated through Midjourney recently won first place in the ‘emerging digital artists’ category at Colorado State Fair’s annual art competition. If that doesn’t convince you, look at these cats:
Now, before you fire your graphic designer or cancel your Adobe Stock subscription, its worth noting that the current generation of AIs are not perfect for every brief. While this may quickly change as the technology becomes more sophisticated, they currently often struggle with very specific details such as human faces or the correct layout of a ‘qwerty’ keyboard in ‘photorealistic’ images.
Similarly, while you can use an AI to generate a new logo for your business, you will still need a graphic designer to clean it up and make the finishing touches. Hence, the key takeaway here is that image generating AIs are not quite ready to replace anyone, however, they can be used as an extremely efficient tool to boost productivity and speed up the development of projects.
One final consideration to make before you jump on board, is to do some research into the datasets used to train the AI that you would like to use. Critics of the technology have accused some companies of training their AI using the work of artists without their consent, which may lead to copyright infringement. In response to this, Shutterstock have announced a ‘contributor fund’ to compensate artists when their images are used to train AIs. If more image hosting companies like Shutterstock follow suit, this should not be an issue in the future.