A question that many creatives has been asking themselves lately is if AI generated imagery will replace our jobs. Dall-E, specifically, is one of the most impressive visual generators that can take a simple prompt and make something well crafted and unique. But can we utilize Dall-E for graphic design or is it destined to take over our jobs entirely?
From what I’ve seen, for graphic designers anyway, there doesn’t seem to be too much cause for concern quite yet. Dall-E creates illustrations in a variety of styles, but it cannot create a logo or layout, and as of right now, doesn’t seem to be great at handling any type of text. Could this change in the future? Absolutely.
At this point in time, using Dall-E for graphic design rather than instead of graphic design seems more realistic. In other words, Dall-E is more of a helper to designers than a competitor.
I got access to Dall-E a few weeks ago and since then I’ve been testing out a bunch of random prompts to see what I can make. I wanted to know exactly how we can use Dall-E For Graphic Design. My prompts were ‘Mario riding Yoshi’, a ‘mountain under a nebula’, a ‘panda wearing a top hat’, and ‘Voldemort holding a Slushee riding a unicorn’. All totally normal stuff.
From those prompts I got a mixed bag of digital illustrations, all with varying quality. Each set had at least one image that could be usable for a design project, either as an image to accompany an article, an image for a poster, or something else.
Then I wanted to try something different; something to test if Dall-E can replace stock photos or an actual photographer. I used Dall-E to create a realistic close-up shot of a woman with red curly hair and freckles. Those results were actually very, very impressive.
From what it seems, at least for graphic designers, Dall-E simply enhances our image search capabilities via original image generation, and competes more with stock photo sites than it does with us. That being said, there is some obvious competition here for photographers and illustrators, and certainly even more to come in the future since this technology is still so new.
Dall-E Instead of Stock Photos
So next I wanted to test to see if Dall-E could truly replace a stock photo. I figured I should create an image prompt for something that I would actually use in a client project. I went back through some old projects that used stock photos to see if I could create something better that what I actually used at the time.
Years ago I worked for a magazine that covered topics like interior design, architecture, and home gardens. I would often need to find a stock photo to use for an article or ad if none were supplied by the author or ad client. Would I have be able to use Dall-E to generate custom images that are better or more unique than a stock photo?
First I gave Dall-E the prompt of “modern home architecture interior kitchen minimalist realistic” to see what it came up with. Now I know my prompt was pretty generic, and if I had a more specific vision in my head of what to create I might have received better results. The first results here aren’t too bad, but there are obviously renders with odd reflections and weird layouts.
Attempting some exterior shots didn’t go very well either. I got two semi-dilapidated homes, one confusing looking exterior, and a somewhat decent render with some texturing issues. In both cases it could just be a matter of trying the right, highly specific prompt, and generating it over and over again until I got something I could actually use for a client.
In the third attempt I tried to generate a living room with a gothic decor style. Only one of the results really turned out ‘gothic’, but even it had issues.
So what if I needed to make an advertisement for a client who had a specific type of photo in mind but no budget for a photographer? Dall-E has limitations, but what are they exactly? I wanted to create a realistic image of a woman with pink curly hair wearing a dress at the beach.
The first batch of images weren’t great, and most came back as illustrations. Then I added the word ‘close up’ and the results were much better. Dall-E clearly has a few issues with creating faces, but overall it did pretty good. You might even be able to Photoshop an image to make it perfect if Dall-E can’t quite get there, or use the new Image Editing Tool, which is still in beta.
The Dall-E Image Editing Tool
The last thing I wanted to to was test the Image Editing Tool in Dall-E. It lets you erase part of an image to edit it or add a generation frame to extend the image. Essentially, it allows you to make minor fixes to imagery that you’ve generated, in addition to expanding it out. It sounds similar to how you use the content-fill tool in Photoshop, but better. Let’s see how it actually performs.
So I chose to edit the highest quality image that Dall-E has created for me so far; the first image created from the close up face of a red head woman with freckles prompt. I generated a little bit out of the image at a time, counting on the AI to fill in the image appropriately. It did pretty well for awhile, but things got weird when it had to generate too much beyond the original image.
The Image Editing Tool didn’t really do a great job here, but it’s still in beta. If I wanted to use this image for an actual project I would probably need to add some better hair and make edits in Photoshop rather than continue along in Dall-E. But still, the initial image is really great.
I wanted to try the Image Editing Tool with one more image; but with an image that wasn’t realistic. I went with Voldement, his Slushee, and the unicorn. I think that particular image has the most potential to create something totally unique and ridiculous.
That it did.
I picked the first image of Voldemort and his Slushee that Dall-E gave, me, then went a bit nuts with it. Basically, I kept asking Dall-E to re-generate different versions of Voldemort holding a Slushee riding a unicorn all over the place, but still adding to the original image. Eventually I generated what I consider to be the greatest image ever created of He Who Must Not Be Named. I’m sure you’ll agree. This is the true power of AI generated imagery.
Dall-E Versus Designer
If you’re interested in seeing a direct comparison between an AI and a graphic designer, you can. A studio called MKBHD posted a YouTube video where they challenged their graphic designer to create prompts that were also given to Dall-E. It’s a great video and you should definitely check it out. The results certainly proved that the designer’s work was of a higher caliber than the AI created imagery. Although, the AI images are created in seconds, rather than hours. Maybe a mix of both AI created art with a designer’s touch is ultimately where we’ll end up?
Dall-E For Graphic Design – In Summary
While it might take a bit of tweaking to get an image that fits what you’re looking for, as long as you’re not super picky about the exact type of imagery you need, Dall-E looks like a great option to generate unique imagery for design work. Fortunately, Dall-E likely won’t replace the job of graphic designers for awhile, but it could replace certain image making tasks that would be done by designers, illustrators, 3D artists, and photographers, but only to a point.
Dall-E is impressive but has a long way to go before it really threatens the design industry. For now, we can use Dall-E for graphic design to create unique imagery, rework our own images, extend an image beyond its original dimensions, and be inspired by the general creativeness of the AI. At some point I expect photo editing software like Photoshop will have this type of AI image generation built right into the program. There is just so much we can do with it.
Have you had any luck using Dall-E for graphic design? Have you used Dall-E created imagery for client projects? Can you think of anytime you’ll likely use Dall-E for graphic design projects? Let me know in the comments below!
I wrote a post awhile back about the app Wombo Dream. It was my first time using AI generated art and I was really impressed. It can’t quite compare to Dall-E for graphic design projects that require any amount of realism, but for abstract art it’s pretty cool. You can check it out here: Wombo Dream, AI Created Art – Is This The Future of Art?