November 12, 2023
I’ve been wanting to do this for a while, but it kept getting pushed to the back burner. I kept wondering, is it possible to do real design work with Midjourney? Is it possible to design all the assets and branding guidelines for a fictional brand using mostly generative AI tools? And could I knock it out over the course of a weekend? Any longer than that, and it would never get finished. It doesn’t need to look perfect, just enough for a proof of concept. I just needed to set some strict ground rules, start with a creative brief, and add a firm deadline. The rules were as follows:
• ChatGPT would write the brief (based on my prompt).
• I could use Dalle-3 for logo ideas, but then vectorize them and modify.
• I would make all images in Midjourney, but could modify in Photoshop.
• I would use Figma for all typography since no AI tool does type yet.
• I would allow a maximum of 10 hours over a 3 day weekend.
• Everything ships at 5pm on Sunday.
I set some loose parameters and asked GPT to write a creative brief based on the following prompt.
You are a creative director. You are writing a creative brief for a new technology company that makes AI Powered Creative Software similar to Canva or Adobe. Please come up with 3 possible brand names and then a creative brief that describes the look and feel of the brand, adjectives that describe it, and communication objectives for what we want our viewers to feel . Primary demographic is creative office workers, aged 20–30, international, major urban centers.
I took the first response as my marching orders and got to work. Here are the three brand names ChatGPT suggested, and why.
CreativIQ: Combining “Creative” with “IQ,” this name suggests a smart, intuitive approach to design, appealing to the tech-savvy and forward-thinking demographic.
VisionaryCanvas: This name evokes the idea of a limitless platform where creative visions are brought to life, catering to the imaginative and ambitious young professionals.
InnoDesign: A blend of “Innovation” and “Design,” this name speaks to those who are constantly seeking cutting-edge solutions in their creative pursuits.
OK, not amazing, but I didn’t come here to nitpick, so I went with the first one, CreativIQ. I received a lengthy creative brief, but I’ll summarize it here:
Visual Aesthetic: Modern, sleek, and user-friendly. The design should be minimalistic with bold accents to signify creativity and innovation. The use of vibrant colors combined with neutral tones will give a fresh and energetic vibe.
Typography: Clean, sans-serif fonts that are easy to read, yet contemporary. This will ensure the brand feels accessible and professional.
Adjectives: Innovative, Intuitive, Dynamic, User-Friendly, Inspirational, Trendsetting
Communication Objectives:
Inspire Creativity: Communicate that CreativIQ is not just a tool but a partner in creative expression.
Empowerment: Show how the software empowers users to bring their visions to life, regardless of their technical skill level.
Community and Collaboration: Emphasize the platform’s ability to connect users from different backgrounds and skill levels.
Cutting-Edge Technology: Highlight the AI-powered features that set the software apart from traditional design tools.
Global Appeal: Reflect the international nature of the target demographic in the imagery and language used, showcasing diverse cultures and ideas.
While ChatGPT had given me some original adjectives, (Innovative, Intuitive, Dynamic, User-Friendly, Inspirational, Trendsetting) I took out my trusty brand deck and threw some additional human labor into the mix.
In my role as a professor, I’m a big fan of these cards. I managed to whittle that down to this: Imaginative, vibrant, optimistic, multifaceted, and energetic.
Based on what I had distilled from the brand deck exercise, I input the following into ChatGPT
Thanks, can you give me a prompt that I can use in Dalle-3 for a logo for CreativeIQ? I want it to use the text “CreativeIQ” as well as pull from the brief you wrote and include the additional attributes, “Energetic, optimistic, vibrant, imaginative, and multifaceted”
The resulting prompt was pretty long and just repeats a lot of the information from the brief, so I won’t paste it here. Instead, let’s look at the results:
I probably should have just spent a hour remaking these from scratch, but instead I wasted an hour chasing individual vector points around, trying to hold onto what Dalle-3 gave me. Then I remembered, ‘This is just an experiment. It’s for the blog post. It doesn’t have to be perfect’. So in the interests of brevity, I simplified the one on the left, and cleaned up the one on the right. Lesson learned, Dalle-3 can help in brainstorming, but cannot make a good logo yet.
The I reread my creative brief, ‘minimalistic with bold accents’, and remembered that AI is supposed to augment, not dictate, my workflow and human input still matters. The logo just worked better when I simplified it. Yes, I could have done this without AI, but it’s an experiment.
Now with a clearer direction, I spent about 10 minutes in Midjourney producing nothing of value before entering this prompt: vibrant young office worker, alternative, hipster, ethnically diverse, editorial photo, volumetric lighting — ar 16:9 — s 50, which produced this:
After I got the hang of it, it was pretty easy to generate about 50 images in a variety of contexts, and then narrow it down to 20. My only complaint is that Midjourney always gave me men, unless I specified ‘female’. While these are cherry picked (below) Midjourney did a really great job at consistent art direction. I always fall into a vortex when I’m generating images and I have no concept of the passage of time. However, today I was careful and clocked it at 60 minutes. However, this only generated pictures of people. I would still need to return for more.
This was a total failure. I tried several different approaches for ‘graphic design software icons’ and nothing even came close. However, Dalle-3 got it right on the first try. When I asked for four more in the same style, the results weren’t quite as good, but enough for me to get started if I wanted to clean them up in Illustrator.
Back to GPT for some more content strategy, and guidelines on voice and tone. Back to Midjourney for some images, Photoshop for tinkering, but mostly making simple layouts from the assembled parts.
ChatGPT was excellent in writing a Creative Brief and helping with content strategy.
Dalle-3 is good at generating rough ideas for logos and icons. This is a bit of a time saver when creating concepts, but not a substitute for rolling up your sleeves and getting to work.
Midjourney is amazing at generating images of people, pretty weak on logos and icons.
I get lazy and want to use AI tools for shortcuts. The better approach is to think of myself as doing most of the work, but delegating some tasks and keeping my focus on the bigger picture. There are no shortcuts.
You can’t make a brand in a weekend. Next time I’ll aim higher, but allow more time.
Total time spent. About 8 hours, over a three day weekend. Not bad for a day’s work. I feel like I’m getting better at my workflow and preparing myself for the road ahead. I’ve read about half a dozen articles in the last week that paraphrase the same Jakob Nielsen quote I put in my newsletter.
“AI isn’t going to steal your job, someone who knows how to use AI will”. My goal with all these projects is to make sure I end up in the second camp.
Since we’re on the subject of my newsletter, why don’t you check it out: https://erikfadiman.substack.com/