AI Essentials: Using Words To Generate Images

Inspired by one of the greats. Generated via DALL-E.

A picture is worth 1000 words.

Now you can use just a few words to create a picture.

Hope you all are ready to get creative!

In today’s AI Essentials, we dive into the exciting world of AI image generators like DALL-E, Canva, Firefly, Titan, and Google Bard!

Before we really start cooking up some images, why would you want to generate images?

Here’s some of the use cases I’ve found customized images helpful for:

1. Visual guides to describe characters, setting, plot, and vocabulary in English and history.
2. Critical visual context builders for real world math and science topics.
3. Customized images to support points in a presentation, blog, or other medium.
4. Assessment and check for understanding modality.

• I actually did this for a recent fellowship I concluded with the National College Access Network. I had the privilege of being a part of a small network of aspiring c-suite college access leaders and we learned about a multitude of different topics including fundraising, branding, change management, and executive skills
• Rather than putting together a power point, I actually generated 5 images which I used as anchors to describe significant moments and learnings during my journey.

Let’s start cooking up some images!

I will be using ChatGPT's image generator DALL-E, which is available to plus, team, and enterprise members and Canva, which is free for educators or 15 bucks a month and includes magic media image generation. All I have to do is enter a prompt like I would in an AI chatbot tool.

Let’s see what images I get back when I enter the following two prompts:

Generate an image of an enthusiastic teacher leading a lesson on math with students.

Generated via DALL-E and Canva

I’m impressed with the subject, tone, and style of the DALL-E image and don’t love the artistry or detail in the Canva one though they also captured the subject accurately.

Let’s say I wanted more control over the specific subject and artistry in the image, I could write a more detailed prompt to get me closer to the vision I had in my mind.

Generate a medium shot, high quality, pop art image of a Latina teacher in her mid 20s wearing modern business casual attire. She has an enthusiastic expression on her face as she reads from a children’s picture book to an overly excited classroom of kindergarten students, who are sitting on a vibrant rainbow colored rug. The classroom is filled with art supplies, toys, and other objects commonly found in a kindergarten classroom. The overall tone of the image reflects the joy students feel in engaging with their passionate teacher.

Generated via DALL-E and Canva

Notice how both images showcase enthusiastic teachers but the second set of images look drastically different from the first because of the specific subject and artistry elements I asked it to include in the generation.

That’s pretty much it!

Including subject and artistic elements should create an image aligned to what you are picturing in your head as you iterate, though just providing a very basic subject prompt could also be sufficient.

A few bonus tips I’ve learned after spending 1000 hours with AI tools like image generators:

1. My favorite prompting words to convey artistry include: words around tone (vibrant, bold, melancholy, somber, etc.), artistry (acrylic, comic book, animated, pop art, laser cut), and camera angle (aerial, extreme close up, medium shot, etc)
2. If you are not specific, this is where you will notice some of the common biases/drawbacks of AI image generators. They are prone to: generate big robots to represent AI, distort human anatomy, default to white males in any success role images, and misspell words unless you are specific about how to spell a word you want included.
3. Just like AI chatbots, you can learn through exploration! I usually approach the brainstorming in two ways:

• I type my high level idea in a text generator like chat gpt, such as I need five thought provoking, engaging image ideas for showing how important collaboration between teachers and parents in supporting students is. I then generate all five images or ask for five new ideas
• I will put in a basic prompt, like our first example, and continue adding elements until it gives me what I want. I usually try to get the subject elements of a picture right first, then I will move on to focusing on the tone, artistry, and style of an image.

Finally, I find this process to be really fun! Try creating images for professional and personal use cases and see what most resonates for you.

Getting started with image generation is as easy as that! Now pick your favorite image generator (I would recommend Canva as it is free for educators. Bard is good too but it is more selective on the images it will generate) and start creating!

For my AI enthusiasts who listened, what are your pro tips around generating high quality images and what do you find yourself creating images for?

Check out the full episode here. Join the conversation at TheAiEducationConversation.com

I’ll be presenting at Sequoia Con 2024!! If your team, school, district, or institution is ready to take the leap in exploring AI implementation, my session will support with strategies for implementation and navigating change management!

The team at Evergreen is also giving any of my followers $50 off conference registration if you use my code: AICONVO.

You can register here. Hope to see you at my session!

#HumansAtTheHeartOfEducation