Before ChatGPT made generative AI accessible to anyone with a keyboard and an internet connection, image generators like Midjourney, DALL-E, and Stable Diffusion were radically impacting the design profession. Launched in 2022, these applications could turn a few words of plain text into richly detailed, often photorealistic images and visual compositions. AI image generation became even more widespread in 2023 when design software giant Adobe released Firefly, a suite of generative AI tools that integrate with flagship applications like Photoshop and Illustrator.
The first wave of Adobe artificial intelligence features focused on image editing. Since then, Adobe has updated Firefly with more powerful generative AI tools that let designers create stylistic variations of an image, watermark their AI-augmented creations with “tamper-evident” metadata, and produce new vector artwork from simple prompts. According to the company, Adobe’s AI-generated art tools have been used to create more than 6.5 billion images as of 2024.
As with all things in the fast-changing field of AI, the rapid rollout of new capabilities can create a learning curve for getting the most out of the tools and using them responsibly. In this Q&A discussion, two Toptal design experts describe the impact that Adobe’s AI has had on their creative processes and productivity. Jessica Souza is a digital designer with a decade of experience creating marketing and branding assets for clients like Google, Lyft, Ferrero, and Spotify. Sérgio Estrella has 20 years of experience as a UI/UX and graphic designer, shaping digital experiences for Fiat, AB InBev, and others.
This conversation has been edited for clarity and length.
To start, which Adobe AI art features do you frequently utilize? Souza: I use Photoshop’s Generative Fill feature daily. I can use Generative Fill and Generative Expand to insert AI-generated imagery into photos or type text prompts to extend an image’s borders. Adobe Illustrator features such as Text to Vector Graphic, which enables you to create vector graphics using descriptive prompts, and Generative Recolor, with which you can change entire color schemes within vector illustrations, are getting better every day.
Estrella: As a UI designer, I find Adobe’s AI doesn’t currently offer as much. They don’t have a dedicated UI design tool anymore since they stopped developing Adobe XD. But I use Adobe Illustrator for my work creating branding guides, logos, and icons—especially now that Illustrator has generative AI vectors.
I use Generative Recolor often—it lets you try different color variations, which is useful for logos and visual identities. Additionally, if I need visuals for a landing page or a feature page I’m designing, I use artificial intelligence in Photoshop to generate images to use as placeholders. Before, I used stock photos and vectors that I could license.
How have the features of Adobe Firefly improved your creative workflow? Souza: When it comes to stock images, clients often want very specific compositions. A customer might, for instance, state, “We need a woman holding a prescription package and her phone in the other hand.” She has to be happy and standing in front of a pharmacy. Her clothing should be casual, with green accents.” I used to do extensive image research, purchase multiple stock images, and spend hours editing them together —and they would never be perfect.
Now I can purchase one image as the base, and then work on it with Photoshop’s Generative Fill. This reduces the time I invest from three or four hours to around an hour, including the image research.
Estrella: AI changed how I deal with icons. Before, you had to search an icon library for the specific object you needed and then transform it into a vector graphic and edit it. Google and Apple both have icon libraries, and there are many others—and depending on what you select, you may need to license it.
By using AI, you can create icons that are more specific to your needs and more authentic. You won’t see the same icon on all your projects, and you don’t need to keep looking at the library of icons—you generate your own.
How accessible are Adobe’s AI art features to novice designers?
Estrella: Adobe did a good job of adapting the UI to make the AI options easy to find without being distracting. If you’re a novice, the AI features are impossible to miss. If nothing is selected in your Adobe Illustrator or Photoshop project, a button for generating an image will appear. If you’re an expert user and actively working on a project, you likely won’t notice the creative tools as much.
Souza: As with most features in Adobe’s suite, the AI art tools are quite easy and intuitive. For example, to use the Generative Fill in Photoshop, all you need to do is select an area of your board and type what you’d like to generate—or don’t type anything and see what Photoshop suggests.
Are there any limitations you’ve encountered when using AI for creative purposes in Adobe software?
Souza: Generative Fill is still developing and doesn’t deliver the same level of creativity and quality as MidJourney. But when I use MidJourney or Pikaso to start the concepts and Photoshop to refine them, then I have the perfect flow.
Estrella: I agree. When it comes to images, sometimes Photoshop’s generative AI is not as convincing. Due to the visual inconsistencies and hallucinations that are frequently found in AI output, editing AI-generated images is also more challenging than editing photographs, especially if the generated image is photorealistic. Instead, you need to keep regenerating or changing your prompt.
When generating vectors in Illustrator, it’s not always as detailed as you want, and it’s not as well-crafted as a human would make. It’s more like a puzzle: the AI doesn’t make a whole object, it makes little pieces that it tries to join together. Sometimes the shapes are not sharp, and it looks like a bitmap image that was transformed into a vector. It doesn’t really matter how much you tweak your prompt—you will always need to edit it.
I wish Adobe offered more AI features that would help UX designers with manual methods and repetitive tasks. Say I’m designing a UI with multiple images, like a gallery or a list of messages with avatars of people—I would like the AI to notice that, and fill in the circles I made with generated placeholders: images of people, their avatars, their names, even the text of their messages. These tasks take time but are not creative work—more like busy work, so it would be great if AI could streamline them.
How do Adobe’s AI-powered suggestions and recommendations help you explore new creative possibilities?
Souza: For me the AI-powered suggestions help so much with photography. I come from an advertising background, working at a time when most brands would have custom photoshoots for their products and campaigns. A good photoshoot has significant costs: the day rate for a studio, photographer, photography assistant, model, makeup artist, art director, and post-production. Now with Adobe’s AI features—sorry, for what I’m about to say photographers—I can purchase a stock image and customize it to a great level, with some limitations, but for a fraction of the cost of a photoshoot. These features are allowing smaller brands to have access to great “custom” photography at a lower cost.
One of my clients is getting compliments on their campaigns from their customers, partially due to the stock images I improved using Photoshop’s AI features. In addition to editing stock images, I also enjoy creating initial image concepts with Adobe Firefly. The image generator includes a reference gallery for both composition and style with images you can add to your prompt, which I often take advantage of. You can also upload your own reference image, but Adobe’s image generator produces excellent results about 90% of the time. Estrella: When you prompt Adobe’s AI with Generative Recolor or Generative Fill, it doesn’t show only one result. It gives you a few alternatives to choose from. This speeds up the ideation process because you can visualize possibilities and see if they’re right for the project before sinking time into them. When you find an option that’s close, then you begin work on editing that image or vector.
On a branding project, my client wanted more color options, so I used Generative Recolor in Illustrator to expand the choices. I actually used the wording that the client sent in a message to me in the prompt. I added more detail to the prompt, but it was basically what the client wrote. This approach generated a color combination closer to what the client wanted and needed, and they ended up choosing one of those AI-generated options.
It made me feel afraid for a moment—like AI was going to replace me. But I see this technology as a tool for making adjustments. It didn’t replace all my work on this branding project; it just helped me adjust one aspect to get closer to what the client envisioned.
