I was originally planning to write this article as a "hands-on with Adobe's new AI features," and cover the new AI features in both Photoshop and Illustrator. As soon as I started to test out Illustrator's new features, however, I realized that there's enough in Illustrator alone to warrant its own story.
I should mention that while I eat and breathe Photoshop on a daily basis and have for decades, I'm merely a dabbler in Illustrator. I use it to set up images for my Glowforge, to tweak some icons and graphics, and to convert raster graphics to vectors on occasion.
Also: Photoshop vs. Midjourney vs. DALL-E 3: Which AI image generator wins my 5 tests?
Unlike the title of the product, I am not an illustrator. I have little experience using the tool to draw stuff. That's relevant for this article because Illustrator can now draw stuff for you. The question is, how useful is that? I've found it's amazing... but it has its limitations.
Let's dive in.
Okay, so you know the drill. You select an area, type a prompt, and the AI gives you a picture. What makes this story unique is that Illustrator gives you vectors, not just pixels.
As with Photoshop, Illustrator has a floating command bar.
Hit Generate Vectors and the tool should open.
Say that we want a deer that likes to visit the apple trees in the backyard on very hot days. I turned that into a graphic using the prompt "Deer sitting on grass under a small grouping of apple trees."
This wasn't quite right, because our deer should be a doe. So I replaced "deer" in the prompt with "doe" and got this.
There's a lot you can do with that, including changing the color scheme and using a reference style. Still, for someone who doesn't draw, this is astonishing enough. For example, Illustrator generated a vector illustration, so you can ungroup the elements.
You can also scale the elements and they won't lose any resolution. As you can see, I resized the doe and it didn't lose any resolution.
Unfortunately, there are a few limitations, both in the AI functionality and in Adobe's implementation of text-to-image. I copied the deer from the image and put it on its own pasteboard. I also dropped a green rectangle behind it so you can see the illustration more clearly.
Notice there are some missing details. At (1) there is no delineation for the chest of the doe. There does appear to be some below at (2), but that's actually provided by the grass. If you remove the grass, the chest definition goes away. Likewise, up on her ear, at (3), the branch and the ear are all part of one object. I copied and pasted it at (4) so you can see how it's all one defined part.
So while Illustrator's AI feature did generate a vector image, it didn't create all the vectors logically, so that someone who doesn't use Illustrator could easily manipulate the image. For someone with even a modicum of experience, though, drawing the curves to complete the chest and remove the branch should take a matter of minutes.
The other limitation is in the AI's knowledge. Because Adobe trains its AI on its own licensed image database, the AI tends to run into some snags. For example, when I gave Illustrator the prompt "Yorkie poo puppy," which is the breed of my beloved little dog, Illustrator gave me this message.
I'm guessing "poo" (which stands for poodle) is just not permitted, even though Yorkie poo is a popular breed. I replaced the prompt with "Yorkie puppy" and got the cute little guy you see at the top of this article.
Next up, Illustrator can now apply generative fill to shapes, using awareness of the shapes as part of the result. Illustrator comes with a demo image where an outline of a bird is filled with stylized feathers and flowers.
But this is ZDNET, so we're going to punch it up a notch. I decided to start with an outline of the ZDNET logo, in a box.
Next, I drew a selection rectangle around the whole thing and gave it the prompt "robot city." It gave me three results, but I liked the first one best.
Illustrator also allows shape fill with a variety of pre-defined styles. Here's what Isometric produced.
One other slider option allows you to increase or decrease shape strength and detail. Shape strength tells Illustrator how closely to stay tied to the shape you're using. For the ZDNET letters, anything other than a shape strength of full resulted in a fairly mediocre creation.
This example showed that Illustrator's AI feature could recognize an illustration. So, using my copious illustration skills, I drew a da Vinci quality illustration as a test.
I didn't give it any prompt. I just clicked Gen Shape Fill, and it clearly recognized my shape as a snowman.
Finally, in the hopes that Illustrator's AI could take my very rudimentary drawing and turn it into something better, I set detail to high, kept shape strength also high, set effects to "minimal," and gave it the prompt "winter wonderland." I have to say, I was pretty impressed.
Finally, we'll take a look at Illustrator's new tool that generates perfectly repeating patterns. One of Adobe's demonstration videos shows how this could be used to generate patterns for fabric, using a service like Spoonflower. I would imagine it would also apply to wallpaper, puzzles, or anything where a repeating pattern is used.
You get to the Generate Patterns feature by selecting its option under the Illustrator Windows menu. I'm starting with a square, and I'll give it a prompt of "brightly colored robots in a 1950s comic book art style." Here's the first result.
I tried a variety of pattern color styles, but I didn't get anything particularly exciting. Here's the best I got using those options.
Then, under Edit => Edit Colors, I found Generative Recolor.
I tried Trippy Disco and got this color combination.
As you can see, the pattern does repeat. To make it repeat, all you need to do is drag a rectangle, because Illustrator sets the current pattern to the one you just created. You could, if you wanted to, make that pattern into a swatch to use later using the Illustrator Swatch pallet.
And, of course, because it's a vector illustration, it's completely editable.
Finally, I went back to the original idea of the doe in our backyard, and turned that into a repeating pattern using "Doe sitting on grass under a small grouping of apple trees" and the Flat Design style:
As you can see, these are some powerful new tools. It will take some time for artists and designers to decide how to adopt them into their workflow, and there is still the issue of copyright and ownership. Spoonflower, for example, requires you to state that you own the copyright for your designs -- but is a pattern you generated with Illustrator yours?
It's not clear that US copyright code will allow these designs to be copyrighted. According to the Compendium of the US Copyright Office Practices, Third Edition, "To qualify as a work of 'authorship' a work must be created by a human being... Works that do not satisfy this requirement are not copyrightable."
Also: Who owns the code? If ChatGPT's AI helps write your app, does it still belong to you?
We'll have to see how this plays out over time. I didn't draw that deer pattern, but I did specify the prompt, so it might be something I could submit to, say, Spoonflower and fairly confidently consider myself the copyright holder. Then again, it might not. I can think of compelling arguments for both sides of the issue.
On the other hand, the robots were fairly generic, and I'd be hard-pressed to say there's anything unique enough about their design or presentation to consider them something I could say I authored.
What do you think? Are you likely to use these features? Do you use other generative text-to-image AI capabilities in other tools? What about the authorship/copyright issue? Do you think that something you generated using an AI tool is something you created? Let us know in the comments below.
