Wave Your Magic AI Wand

Use Generative AI Image Creation to WOW Your Fans

“From Bella’s Point of View”

Wave Your Magic AI Wand

Use Generative AI Image Creation to WOW Your Fans

“From Bella’s Point of View”

All images in this post are AI-generated

The AI explosion

They were everywhere! I kept seeing more and more AI-generated, “fake” Disney movie posters in Instagram. Every time I scrolled down I saw another adorable pet dachshund “starring” in an upcoming movie, or appearing in some exciting travel location—all shown in Pixar-animated movie drawing style.

I had to figure it out tout de suite! The implications of AI-generated images for future graphic design and online imagery are huge.

Sure, it’s fun. They’re cute. But what does any of this have to do with marketing your business, or building a website?

All images in this post are AI-generated

Why AI images are important to your business

To grow your business, you have to increase your visibility! The more visibility you have, the likelier you are to pull prospects into the marketing funnel and gain additional clients. Thus, great images are front and center to gain that visibility and disrupt doom scrolling.

In recent years, we’ve seen enormous changes in the image/video industry. Professional photographers have become increasingly squeezed out as the business model has shifted dramatically. Business owners often can’t afford to generate all their own high-quality photos and videos for websites and social media. So they turned to stock photo/video providers, which have in turn shifted to subscription plans to meet the increasing demand for high volume imagery. And now consumers become suspicious when they spot over-used stock imagery by businesses.

Why should you consider using AI-generated images? The possibilities are endless, and are far from “stock.” But the potential for abuse is high. You don’t want to use an AI-image while simultaneously claiming it’s a pro photographer’s work. Being transparent with prospects is important to gain their trust. Here are few reasons to consider using generative AI imagery:

  • The image you envision using can’t be taken here physically. (Perhaps you want to put a beach scene on top of Mt. Everest?)
  • Your products are in prototyping, but you need a simulation of them to photograph and begin building interest in them.
  • Cost of production is prohibitive. (That enormous crowd of extras clamoring at you on the red carpet has an expensive price tag to pay for their time.)
  • You’d like to capture that “Barbie the Movie” color scheme look.
  • Van Gogh is unavailable but you’d love to use the style of his paintings to set the mood on your website.

How do AI-generated images fit into your business model? Does it best meet budgetary constraints? Is it the best way to match your brand styling? Does it give you the most creative control?

What imagery is best-suited for generative AI?

It doesn’t matter if the image has a dachshund or your own face in it: what imagery will grab your prospects’ attention?

I can’t tell you the number of times I’ve heard from clients that they wish they’d had more dynamic images for their project. Maybe the image they have is too small, cropped too tightly or oriented incorrectly. Maybe it has the wrong background, or it’s just plain boring. What can they do?

That’s where AI-generated imagery may play a role in your business. And the evidence is there this year in the AI releases and clamor for these tools. Even the Collins Dictionary has chosen “AI” — artificial intelligence — as its word of the year. Recently, this AI technology has made another leap that may aid your business (not to mention provide some fun): DALL·E 3, the AI software that generates images “out of thin air,” is now available for free via Bing or ChatGPT.

Intro to creating DALL·E 3 images

What DALL·E 3 images can you create? Where can you use them? How do you make them? Is it legal?

First, AI-generated images are never a substitute for great, professional photography. Whenever possible, please hire a photographer for your business’ team headshots, for product and service imagery, and to obtain the most unique visual representations of your offerings.

Second, use AI to enhance the pro photographer’s images in ways they can’t accomplish. (More on this below.)

DALL·E 3 can create images of whatever you can imagine. BUT, the software creators have baked in safety rules and limitations. Reportedly, it will not generate images of violence, abuse, other gratuitous, negative imagery to further bolster fake news or infringe others’ copyrights. It also won’t generate imagery that mimics artwork of currently living artists (to protect rights to their work and earnings).

Created images may be downloaded for you to use on social media, personal use and websites.

It is legal from Bing and ChatGPT, but I would strongly advise reading the terms of use and captioning AI-generated images to demonstrate your transparency. And, of course, the content of what you create can’t infringe upon others’ copyright protections.

How to create AI-generated images

AI-generated images are created with text prompts. You describe what you want to create and hit “Generate.” The software interprets your text, and 30+ seconds later it displays variations of your prompt request. Let me walk you through a few variations I tried, beginning with this image of a dachshund:

Bella's point of view, square, on sidewalk

Bella's Point of View

Prompt: Disney Pixar poster titled "Bella's Point of View" of older smooth red dachshund with white face and pink halter leash walking on sidewalk, smiling up at lots of people's legs (from the knees down) as they walk by her

That was not my first attempt to generate “Bella’s Point of View.” Like all new skills, my earlier attempts produced a younger dog. (Bella’s a senior dog.) The software attempted to flatten Bella into the sidewalk, a la Mary Poppins and her sidewalk chalk art. It may take several adjustments to your text prompt to generate what you seek.

Creating AI-generated photo-realistic images

Photo-realism is more challenging to create. It can take more practice to provide viable prompts. One thing I like to try is juxtaposing visuals that ordinarily don’t go together. They are more likely to get viewers to stop to figure out why dissparate items are combined. For my first attempt, I wanted to try adding 19th century Sherlock Holmes to a modern, 21st century home:

Sherlock Holmes with Dr. Watson in white decorated room

Sherlock & Watson in the 21st Century

Prompt: a realistic photo of detective Sherlock Holmes, wearing brown tweed deerstalker cap and cape, searching intently with a magnifying glass for clues in the all-white living room of an ultra modern house interior with minimalist furnishings; and Dr. Watson, dressed in 1890s suit, looking at living room surprised

Sherlock in foreground with magnifying glass
Benedict at Sherlock foreground
Sherlock Holmes alone in room

As you can see, some variations are more successful than others. The juxtaposed imagery may be more successful if I use color in the 21st century objects and prompt Sherlock and Dr. Watson to be represented in black and white. You also notice AI has difficulty properly generating eyes. (And it seems to think Sherlock should look like Benedict Cumberbatch instead of Jeremy Brett, Basil Rathborne, or Robert Downey, Jr.)

Here’s another attempt at realism combined with a visitor you wouldn’t expect in the scene:

Closet with alligator

Taming the Closet

Prompt: A realistic wide shot of a darkly lit large walk-in closet displaying shoe shelving, drawers, shirts or blouses in double-hanging, and a center island of drawers made from ironwood, with large alligator walking through the closet

Although the above image appears fairly realistic, I found DALL·E 3’s understanding of alligator anatomy to be poor. Look at the variations’ legs on the animal!

Gator with tale on island
Gator with long legs
Gator in back of closet

Nevertheless, while DALL·E 3 may have many shortcomings, it is vastly less expensive as a tool than hiring an animal wrangler and all the other accoutrement to create outrageous imagery that will grab attention.

Generating your own AI images

You can make your own! Here’s a bit of background: DALL·E is an image generation AI model created by OpenAI. (It was first launched in January 2021, with the latest release being its third iteration.) The software app is free to use at Bing. (If you don’t have an account associated with Microsoft, which produces Bing, it’ll ask you to create one.) It’s also available for users of ChatGPT Plus, a new pay subscription version of the free AI software. Resources and links for detailed instructions are below.

In addition to DALL·E, Stable Diffusion, and Midjourney are competing apps for generating AI imagery. Many users combine images on multiple AI apps for image generation as each has strengths and limitations. For example, you may have noticed all the AI-images shown in the story are square-shaped, perfect for Instagram and other social media platforms. But none of these images have been portrait or landscape oriented.

But wait! The image at the top in the header (and shown below) is panoramic. Where did you get that?

Bella's point of view, square, on sidewalk
AI Generated cartoon image of dachshund on sidewalk

Going beyond DALL·E 3

DALL·E’s not the only game in town! I mentioned a few of the primary competitors to OpenAI’s image generation app above, but I do have an ace up my sleeve: Photoshop. With its 25th release of Photoshop this past month, Adobe incorporated their own AI-generative image app, Firefly.

With a bit of trial and error in Photoshop, I was able to nearly triple the width of my original, square image generated by Bing. And just like the competitors’, Adobe’s software offers several result variations from which to choose. (I liked the one with red sneakers best.) Further, I can then add layers to smooth out edges or alter coloring.

Photoshop’s AI image generative tools are a HUGE game-changer:

  • You can edit existing images (which you can’t do on Bing or ChatGPT).
  • Have a photo bomber in an otherwise perfect image? I can remove them with ease!
  • Just want to change one part of a DALL·E’s AI image? I gave that a try with the alligator in the closet (below).

Here’s where I discovered some of Adobe’s first shortcomings. I attempted, through different prompts, to convince the AI “genie” to shorten the gator’s legs to appear more realistic. These were the variations it returned:

Standing gator in closet
cartoon dinosaur-y gator in closet
Stool at gator in closet

Aren’t they pathetic? A stool? Come on!

I’m betting if I worked on it long enough I could get the gator fixed. Fortunately, Photoshop has a tool to report errors or unacceptable results. This will help the programmers teach the next iteration of the tool to produce better images.

On the positive side, someone else produced the nice cornucopia image below on Bing. (I couldn’t resist borrowing it.)

Cornucopia with bounty of fruit

But I didn’t want a square image for using in my newsletter, I wanted one at 16:9 aspect ratio or similar.

I cannot tell you how many hundreds of images I’ve edited in Photoshop over the decades where I’ve manually added “imagery” beyond the edges of the cropped image to make the final image have a higher ceiling, or more flooring (hardwood floors are a bear to recreate) or longer walls. Now, I can just describe what to add and voila! It fills it in. Here’s the expanded cornucopia image the Generative Fill tool created in Photoshop:

Cornucopia with bounty of fruit with curtains

Is generative AI image creation easy?

Yes and no. It’s much easier and faster than creating walls and floors from nothing. But it takes practice to do it well and an eagle eye to spot errors. You may have noted in the wide cornucopia image above that the wooden table top devolves into a tablecloth on its side. (Wood doesn’t drape like fabric! There are other generation errors in other images above I haven’t pointed out.) But you make the choice about how perfect the image needs to be for the purpose at hand.

When first doodling with DALL·E in Bing, I experienced issues with the software hanging up and freezing, misunderstandings with the text prompts and so on. Like anything new, it takes practice.

Putting AI image generation to use

Is there an image you’ve wanted to use on your website that has a problem in one area? Perhaps we can “erase” it. Loathe the wastebasket that accidentally got left in an image? We can omit it or make it into a Ming vase. Have an idea for adding a dash of humor to your work? I dashed this crazy idea off in thirty seconds:

Sunday in the Park with Dachshunds

Sunday in the Park with Dachshunds

Prompt: replace Seurat's woman in the "Sunday in the Park" with a female dachshund

George Seurat must be turning in his grave…  Time to get creative!