Much has been written about creativity and Artificial Intelligence (AI) tools over the past year or so, the majority of which has focussed on OpenAI’s ChatGPT tool. While the ‘Large Language Model’ of ChatGPT is very impressive (and only seems to be growing in usefulness), it is not the generative AI tool that interests me the most as a ‘creative professional’ (yuck)!
Before I get into why I want to spotlight these tools, a caveat: I am not going to touch on the ethics surrounding these tools. Nor am I qualified to explain in any detail how they work. Finally, this is not a recommendation of any particular tool – merely my experience and opinion on using them.
Also, this is a rapidly developing sector, so this info may get out of date very quickly!
A text to image ‘generative image creator’ that develops pictures based on a written prompt.
I believe Midjourney is the most interesting and ‘advanced’ image generation tool currently available. It is able to create a staggering variety of image styles (from incredibly photorealistic pictures, to works of pure fantasy) at a very high quality, in seconds. The rate at which it has developed has been mindblowing too. It has gone from a platform where certain aspects were widely ridiculed (note this link shows a picture of distorted hands, please do not click if you may find that upsetting), to one where many of the more realistic pictures cause me to double take.
I have used Midjourney personally and professionally to quickly prototype ideas and to get inspiration. Because the text to image generation isn’t perfect, I find that Midjourney will serve me at least one image per prompt in a style or composition that I would never have thought of. In this link you can see a showcase of Midjourney artwork. (Please note that, although Midjourney has strict controls over the kind of images it can generate, some images can come quite close to the line of ‘Not Safe for Work’).
Similar to Midjourney in that it creates images from text prompts. However, there are some interesting twists that set it apart from Midjourney.
MidJourney is very focused on creating the ‘best final artwork’ possible from the text prompts you enter. Firefly, however, seems to set its stall out as more of a tool for designers to collaborate with. The current version of Firefly feels like you are being served an inspiration or starting point for an image and there is an unwritten expectation that you, as a designer, need to put in the effort to create the final artwork. And honestly, that’s a good thing. As someone who works with Adobe design tools, I have very specific requirements and am normally after a very specific look, which no text to image generator could ever understand.
Other aspects of Firefly that make it stand out (and a more useful tool) are the fact that it can generate text. It can’t write copy for you (go to ChatGPT for that 😉), but it can create text for design layouts, in a range of really interesting styles.
The final advantages it has is in its integration with Adobe products. At the time of writing several design tools have Beta versions that use Firefly in clever ways. Generative Fill has captured the most headlines – being able to expand or replace things in a scene using a text prompt. But the AI colour palettes is also a game changer for the kind of work I do.
As AI continues to evolve the future could hold even greater promise for AI-powered creativity. As technology advances, we can expect to witness further integration of AI tools into the creative process, fostering innovation and pushing the boundaries of artistic expression. It is certainly a very interesting and exciting time to be a creator in the digital space.