Visual Brainstorming with DALL•E 3 and ChatGPT

DALL·E 3 is now available in ChatGPT Plus

This is one of the most powerful updates to ChatGPT yet.

DALL·E 3 is OpenAI’s newest text-to-image model. Unlike DALL·E 2, it’s integrated within ChatGPT rather than being an standalone product. This minor difference has a significant impact because we can finally combine the power of ChatGPT and use its outputs to generate a variety of visual assets.

Let me show you through an experiment I tried out this week.

Designing a ridiculous product for Formula 1

I watch Formula 1 a lot, and I always joke with my friends about how they needed a giant Roomba-style vacuum to quickly clean up the tracks when a crash happens.

Crashes can take a few laps to clean up, and nobody wants to watch the world’s fastest cars driving slowly behind a safety car until all the debris is cleared.

So I thought this would be a fun experiment to test out DALL•E 3 and ChatGPT’s capabilities to generate:

1. Designs for the giant Roomba-style vacuum
2. Logos for the company that would manufacture them
3. Storyboards that describe the problem and demonstrate the solution

Spoiler: I was able to do all of that.

#1: Generating designs for the vacuum

I started by giving ChatGPT some context about my idea and the problem I was trying to solve.

Me: Alright... I watch Formula 1 a lot, and one of the biggest problems and time wasters is when the cleaning crew needs to clean up the track after a crash. It ruins the momentum of the race, and people don't want to watch the fastest cars in the world driving slowly behind a safety car until it's all cleaned up

‍ChatGPT: That's an intriguing idea! A giant Roomba-style vacuum designed for racetracks could indeed be a game-changer for rapid track cleanup. Would you like a visual representation of this giant Roomba-style track vacuum or further details on any specific aspect?

‍Me: Yes, please make a visual representation so I can better envision your suggestions!

‍

DALL•E 3 generated four design concepts based on the limited information I provided:

Solid start. The first design was my favourite, but I wanted to adjust the colours.

Me: I love the first one. Can you paint it in red? That'll follow the Formula 1 branding, but also make it easier for drivers to see it on the track.

Much better. I still wanted to add something unique to give it a “racing theme” because it was starting to look like an Iron Man suit-turned-vacuum. So, I told ChatGPT to add some checkered flag patterns:

Me: That looks SICK! Can you include some checkered patterns that look like a checkered racing flag?

Now THAT’S a product I can get behind! Don’t forget to give kudos…

#2: Generating a logo for my new vacuum company

Keep in mind that all my inputs and images so far are within the same conversation, so all that context is preserved for the outputs that follow.

Now that I had a unique product, I needed a company name and logo for my blooming racetrack vacuum empire.

I've never had a great experience generating good names with ChatGPT. All the options it gave me felt too "on the nose".

I decided to use Clean Operator after a few iterations. If you watch F1, it's a reference to Carlos Sainz's Smooth Operator nickname (or the song by Sade, depending on who you ask).

Me: Let's use "Clean Operator" for now, we can always change it later. What would a logo for the company look like? Generate some options

Notice how I didn't specify what the logo should have. Yet, it maintained the racing and vacuum theme because of the prior context. Plus, DALL•E 3's ability to generate text within an image has massively improved, as it generates the company name perfectly in the last design.

The fourth option (bottom-right) was the most appealing to me, but I wanted the vacuum to be part of the logo.

Me: I love the 4th one. Can you generate some options that include the vacuum we designed together? I'll pick my favourite after.

Again, it preserved the red colour and checkered pattern I mentioned at the beginning of the conversation when we initially designed the vacuum. I didn't have to repeat myself. I only had to refer to the initial vacuum design.

However, it didn’t preserve the exact design we had: it went for a traditional-style vacuum rather than a circular Roomba-style one.

After unsuccessful attempts at incorporating the circular vacuum design into the logo, I decided to stick with the third option (bottom-left).

#3: Generating storyboards to describe the problem and solution

Storyboards are a great way of visualizing a problem and demonstrating a solution you’ve designed. They can be used for presentations, pitches, commercials, video demos, and more.

I started by asking ChatGPT to generate the storyboard sequence, without images.

Me: Let's take a step back and brainstorm the story and sequencing we want to follow before generating any images.Of course! Let's craft a compelling narrative for the video that effectively demonstrates the product and its value proposition. Here's a suggested sequence:

Storyboard Sequence:
1. The Excitement of Racing:
Begin with the thrill of a race. Show racing cars zooming by, the intensity of the drivers, and the excitement of the spectators.
This sets the stage and immerses the viewer in the world of motorsport.

2. The Unexpected Incident:
A sudden crash occurs. Debris is scattered all over the track. The music changes to a more somber tone.
Emphasize the gravity of the situation and the potential delay it can cause.

3. Traditional Cleanup:
Showcase how the cleanup is typically done: Marshals rushing to the scene, manually picking up debris, using basic tools. This process is slow and halts the race momentum.
This highlights the problem and sets the stage for the solution.

4. Enter the Clean Operator:
A specialized vehicle approaches, deploying the 'Clean Operator' vacuum. Its vibrant red color stands out, and the checkered pattern gives it a racing vibe.
This moment is the turning point, introducing the hero product.

5. Efficient Operation:
The vacuum gets to work. Showcase its features: picking up debris, absorbing liquids, LED lights guiding its path, and sensors ensuring it avoids obstacles.
Highlight the speed, precision, and efficiency of the product.

6. Back to the Race:
With the track cleared, the race quickly resumes. Cars zoom past the cleaned area without issues.
Emphasize the minimal disruption and the product's effectiveness.

7. Satisfaction & Approval:
Marshals and race officials give approving nods or thumbs-up. Perhaps even some spectators or team members show appreciation.
This reinforces the product's value and acceptance by those in the industry.

8.Concluding Message:
End with a powerful tagline, something like: "Clean Operator - Revolutionizing Track Cleanups. Because the race must go on!"
Leave the viewer with a memorable message and a call to action (e.g., visit the website, pre-order now).

Next, I asked it to create a visual representation for each step:

Me: I love this story progression! Can you create a visual representation of each step?

The storyboards I got (after a handful of iterations):

Making my elevator pitch to F1

Now that I had a clear billion-dollar idea (conservative estimate), I wanted to put it all in a concise and compelling pitch.

I didn’t use AI for this part. I just gathered all the visual assets I generated into an elevator pitch you can view.

View Clean Operator pitch deck

The future of the creative process

This was one of the few times I felt like Tony Stark using JARVIS, even if I was designing an unrealistic and oversized vacuum that will never see the light of day.

Combining the power of text and image generation opens up many possibilities and opportunities for the creative process. The most significant power comes from preserving the context and its continuous enrichment as you keep conversing and iterating.

If I wanted to do this last month, I would need to use multiple AI tools in parallel, manually transferring the context between them and writing overly descriptive prompts to achieve the same results. When a single tool offers me all the capabilities I need, I experience the true power of AI because the same context can be repurposed for different objectives. I can generate a business plan, taglines, logos, product concepts, storyboards, and more.

It finally felt like having an always-available brainstorming partner who can quickly visualize the weird and crazy ideas in your head in a fraction of the time. And it was fun to use.

We've only scratched the surface of visual brainstorming. Imagine a world where we can quickly generate and brainstorm:

• Designs for an app or website using a company’s design system
• Marketing campaign assets with specific taglines and branding
• Concepts for new physical products
• Storyboards for animated movies using specific characters

The possibilities are exciting.

The typical question arises again: does this make designers and creatives irrelevant?

Definitely not.

Maybe I'm biased because I'm a designer, but that's been my stance for all AI applications.

AI is only as good as the person using it. It automates the doing, not the thinking behind a design, product, or brand. It should augment us and make us better at what we already do. The only difference is that we can do it much faster.

Instead of getting bogged down in the minutia of putting visuals together, we can focus on the higher-level thinking and creativity of the process.