About Generative AI
Generative AI is a technology that can create new things, such as images, videos, text, and voice. It learns from a large library of examples, such as stock images or sourced from the internet. What’s amazing with it today is the quality of content that it can produce, in various styles.
With prompt engineering, which means giving more specified directions to the algorithm on what to do, you can steer its outputs. You don’t need any technical skills today to prompt images with generative AI, but you can improve your prompting skills by learning techniques in prompt engineering, and how to use AI-assisted tools in software such as Photoshop and Illustrator.
It doesn’t take away the human element though. To generate high-quality and visually pleasing images using AI, you still need taste and knowledge of composition, colors, and various editing tools. However, it does speed up the process immensely if you know what you want and you learn how to direct the AI to produce your desired results! I think it’s simply amazing.
My first experience with generating images using AI
I am currently enrolled in a course in Design and AI at the University of Borås. One of our first tasks was to generate images using Adobe Photoshop, Illustrator, and Firefly using prompt engineering, generative fill, and various style and composition references. I had previous experience using Photoshop and Illustrator, but never really tried to sit down and generate high-quality graphic design pieces before.
I’ve always wanted to create futuristic-looking cyberpunk images, conceptual light art installations, and interior design of my dream homes. I have many times tried to learn 3D software such as Blender, TouchDesigner, and so on, but my skills haven’t gotten quite as good as I wished them to be. In other words, I haven’t been able to create my vision in these areas before this.
Using generative AI though, I have for the first time succeded in creating images that look very similar to the image in my mind’s eye! Look at these results!
Project 1: Light art concept poster
My process for creating this
I wanted to create a poster for a light art installation event, so I wanted it to look like a light art sculpture, photographed professionally. I started with a more general prompt but it was way too abstract. Then I used one of their premade shapes as a reference and it began forming into my vision! However, I wasn’t happy with the colors, so I did a few iterations of that and lowered the intensity of the reference shape.
Tool used: Adobe Firefly
Prompt used: “Design a poster for an art exhibition with abstract shapes with light artistic shadows, the image is blurred and out of sharpness; abstraction. Cyberpunk theme. Light colors”
Additional controls: I added style reference, composition reference, studio lighting, cool toon color, and metal material as references.
Project 2: Loft Interior Design
My process for creating this
I wanted to design the interior design home of my dreams – a harmonious home library in a loft attic environment with plants and slanted ceilings, using generative AI. So, I begin with a loose reference image from Photoshop and my prompt to get the first image. I wasn’t happy so I readjusted the prompt to specify the look of the ceilings, how the books should be, and specifically what plants I wanted. Then I went in with the generative fill for custom-selected areas and fixed details in the room, removed items, and changed colors. I also expanded the ceiling using the generative background tool to give the image a more breathable feel.
Prompt used: I lost my original prompt due to merging the layers. But I believe it was something like “A modern loft apartment with two slanted ceilings and exposed white beams, interior design with monstera plants and lots of stacked books”.
I love it!
My friend and colleague then used a software called Runway AI to generate a motion graphics video of my generated image (PNG format), and this was the result! Pretty crazy
What I love the most
- Generative expansion (it can generate and fill out backgrounds)
- Generative fill (it can remove parts, add objects in selected parts of the image through prompts, and give multiple suggestions)
- You can use reference images to specify what you want. I like this because I think in more visual terms than textual, so it makes my work process easier.
Future projects
Going forward, I want to explore how generative AI can be used for brainstorming UI solutions such as for heading sections, as well as how it can be used to generate image assets for marketing purposes and details such as icons and illustrations for UI design!
Have you tried generative AI for images yet? What do you think about it?