- Illumined Insights
- Posts
- Generative Art with DALL-E 3... Fun? Practical? Both?
Generative Art with DALL-E 3... Fun? Practical? Both?
Use Cases for DALL-E 3
Stephen Hill
November 07, 2023
Welcome to the “Illumined Insights” newsletter! Thank you so much for subscribing. This weekly newsletter touches on all things analytics and data science with a focus on areas such as data visualization, AI, and sports analytics.
This week we take a break from our series on integrating ChatGPT and Excel and shift to a different AI topic: Using AI tools to generate art. I’d love to see what creations you can come up with after reading this week’s edition 😀
Stephen Hill, Ph.D.
DALLE-3 Image
In this week’s edition of the newsletter we take a look at a couple of use cases and tips for the DALL-E 3 text-to-image model. In 2022 OpenAI released its DALL-E 2 model. It’s safe to say that this model (and others like it, such as Midjourney) have revolutionized the digital art creation space. With DALL-E anyone can create digital art by providing a text prompt. For example, this prompt:
Create an image of a futuristic cityscape at dusk, as viewed from a high vantage point. The city is alive with neon lights and holographic advertisements, showcasing a blend of cyberpunk aesthetics and biophilic design. Skyscrapers are covered with lush vertical gardens, and the sky is dotted with flying cars that have sleek, aerodynamic designs. In the foreground, there's a balcony with a clear glass railing, where a robotic owl with glowing eyes is perched, looking out over the city. The scene combines ultra-modern technology with nature, and has a peaceful yet vibrant atmosphere.
results in the image below:
The newest version of DALL-E, DALL-E 3, was recently (October 2023) released and offers improved performance and direct integration with ChatGPT. Note that (as of early November 2023) DALL-E 3 is only available to ChatGPT Plus subscribers ($20/month and well worth it in my opinion). Much of the content of this edition of the newsletter is also applicable to DALLE-2, but DALLE-3 will produce better results.
Ask ChatGPT to Create Your DALL-E Prompts
For our first tip we suggest leveraging ChatGPT (or similar Large Language Models) to help you develop text prompts for DALL-E. If you’re not sure how to construct a text-to-image prompt, it’s completely reasonable to ask ChatGPT for help. For example, I asked ChatGPT to create a prompt that I could use in DALL-E 3 to generate an image to use as the first image in this newsletter post. Here’s the prompt I provided:
Create a prompt that I can use in DALL-E 3 to create a photorealistic image to use as the opening image in a newsletter post entitled "Generative Art with DALL-E 3... Fun? Practical? Both?"
Here’s the prompt that ChatGPT returned as its response:
Create a photorealistic image of an artist's workspace, with a modern computer displaying a colorful and intricate generative art piece created by DALL-E 3 on the screen. The desk is scattered with art supplies like paintbrushes and sketchbooks, suggesting a blend of traditional and digital art methods. In the background, there should be a wall with a playful collage of various generative art pieces, showcasing a mix of abstract and realistic styles. The overall atmosphere should convey creativity, fun, and practicality in the use of AI in art.
Here’s what DALL-E 3 (in ChatGPT) returned to me when I provided this prompt:
DALLE-3/ChatGPT Output
Both of these images are pretty impressive and are far better than anything I could have created on my own. I liked the image on the right a bit better, so I selected that one for the newsletter. I could have also asked DALL-E 3 to iteratively improve the image (we’ll see this in a moment).
Create Stock Images
I absolutely love stock image sites like Unsplash or Adobe Stock, but sometimes I can’t find that perfect stock image to use in a presentation, on a website, etc. Let’s see how well DALL-E 3 does at creating stock imagery. Here’s our first prompt (prompt creation assisted by ChatGPT):
Create a detailed, professional stock image of a group of business professionals engaged in a meeting in a modern conference room. The room should be well-lit with a large conference table, comfortable chairs, and a large screen for presentations. The participants should be dressed in business attire, actively discussing, with some taking notes and others presenting. Include elements like laptops, notepads, coffee mugs, and a whiteboard in the background. Capture the essence of teamwork, communication, and a productive work environment.
Here is the DALL-E 3/ChatGPT response:
Both are quite good. Let’s see if we can improve the images a bit. I’ll focus on the image on the left. Here’s the next prompt:
Let's keep the image on the left, but improve it. Make the image on the board less busy and make the overall atmosphere more light with more white and other soft colors.
Here’s the result:
Not bad. Notice that the image is clearly not just an improved version of the image before. It’s a whole new image, but one that’s similar to its predecessor. Don’t expect DALL-E to be able to fine tune an image as you might do in a photo editor.
Let’s try a different stock image setting using a different style of art:
A hand-drawn cartoon image of an individual working on a laptop in a cozy coffee shop, surrounded by a casual and relaxed atmosphere, with a cup of coffee and a notebook on the table.
Here is the response:
Pretty good. What if we present the same prompt, but change the style?
A flat art cartoon image of an individual working on a laptop in a cozy coffee shop, surrounded by a casual and relaxed atmosphere, with a cup of coffee and a notebook on the table.
I kind of like this style. Part of the fun in working with DALL-E is experimenting with different styles. As a side note, pay careful attention to how text appears in DALL-E generated images. In previous versions of DALL-E, the displayed text was often non-sensical. In DALL-E 3 this seems to be improved, but is far from perfect.
Logo Creation
Every company or organization needs a great logo! Is DALL-E 3 good enough to replace the work of a hired professional? Let’s try to find out. Here’s our first prompt (for a fictional tutoring company):
Create a modern and friendly logo for an online tutoring company named 'EduBright Academy'. Incorporate vibrant colors like blue and green to convey a sense of growth, trust, and learning. The font should be clean and easily readable, with the words 'EduBright Academy' included in the design. Aim for a design that appeals to students of all ages, from children to adults.
And the response:
Not terrible. A bit generic. DALL-E’s tendency to misspell words rears its head here. Out of the two, I kind of like the one on the right the best? What do you think? Let’s try to fix it:
Let's keep the logo on the right. Correct the spelling of "EduBright", place the logo on a white background, and eliminate the glow around the logo's edge.
Not exactly what I had in mind. Still with the misspelling. One more try:
EduBright is still misspelled. Change only the spelling to correct it.
Congrats DALL-E 3! You did it!
After some back-and-forth we’ve ended up with a serviceable, albeit bland logo. The potential is definitely there for DALL-E 3 to generate a reasonable logo. Let’s try one more. This time, a logo for a sports team. We’ll honor the long gone Birmingham Steeldogs (an arena football league team from the early 2000s) by creating a new logo for them:
Create a logo for an American football team called the Birmingham Steeldogs. The logo should feature a fierce, stylized steel greyhound dog with angular, metallic elements to emphasize the 'steel' aspect. Incorporate the colors silver, black, and blue. The greyhound should have a determined, competitive expression and be in a dynamic pose to convey speed and agility. Include the team name 'Birmingham Steeldogs' in a bold, modern font that complements the steel theme.
Pretty good! Except for the “Steeldoogs” in the logo on the left and two “L”s in “Steeldogs” on the right. The logos themselves are pretty impressive. I’d wear either of them on shirt 🙂.
Let's keep the image on the left and only correct the spelling of "Steeldogs" and remove the words "American Football".
Argh! I was so happy about the first results and now we get a misspelling of “Steeldogs”, “Birmingham”, and “American”. Such is life when working with DALL-E 3.
Note: I tried two more times to have the spelling corrected to no avail. I’ll save you the time of seeing the results (the misspellings remained).
Closing Thoughts
I can easily spend a few hours tinkering with DALL-E 3. To think that such a tool is readily available that allows us to create compelling digital art in a moment is staggering. Is the output perfect? No. Is it good? Most of the time, yes. With a little prompt engineering and some patience, a user can generate results that look professional. Should graphic designers fear for their future? Maybe not just yet, but soon.
What will you create with DALL-E 3?
Illumined Insights Book Recommendations
I’m teaching a Sports Analytics Course this semester and recommend a couple of books to my students and to the readers of this newsletter. Note that these are Amazon affiliate links. I may receive a small commission if you use a link to purchase a book.
Feedback?
Did you enjoy this week’s newsletter? Do you have a topic, tool, or technique that you would like to see featured in a future edition? I’d love to hear from you!
Support the Newsletter?
Support this newsletter with a “coffee” (optional, but appreciated).
Start Your Own Newsletter?
This newsletter is created on and distributed via Beehiiv, the world’s best newsletter platform. Want to start your own newsletter? Click below to get started. Please note that this is an affiliate link. I may receive a small commission if you sign up for Beehiiv via this link.