Use generative AI imaging tools for a creative marketing campaign.
If you have been creating images with an AI text-to-image generator like DALL-E 3, Midjourney, Stable Diffusion, or Adobe Firefly, you are part of an unprecedented influx of GenAI digital images born into the world since July 2022.
According to statistics shared by Everypixel, in the first year that AI text-to-image generators became public, people generated over 15 billion images.
15 billion (and growing) is a big number, but not a surprising one. The ease of typing a short phrase and seconds later receiving a composed image in any imaginable style is an addicting prospect.
I am guilty of creating my share of throw away images using AI image generators. I rarely think about the data junk I am producing.
Let’s pause a moment to contemplate about the cost of storing all those images.
If we assume an average image size of ~ 1mb (1024 X 1024 pixels), that means 15 billion images requires ~15 petabytes of storage. A conservative estimate of the cost required to maintain this scale of data comes in at around $4M/year.
I am sad to share that I am still using up kilobytes to save my first text-to-image creation. I am keeping the picture for posterity’s sake (see below).
As you can tell, this image is not a winning racehorse.
To show my progress from this initial generative text-to-image attempt, I wanted to share one way I am using generative AI to create images that serve a purpose.
I hope this example will inspire more thoughtful collaboration with a favorite AI bot. Shouldn’t we generate more purposeful pixels to live on the server farm?
Start with a brief
While my day job is to perform creative work for others, I am an avid tree lover. I enjoy thinking of ways to bring more awareness about the importance of trees to our collective well-being.
I decided to use generative AI to help design a custom image that a nature conservancy might use for a new tree planting email marketing campaign.
To kick-off the project, I wrote a full brief with a little help from Chat GPT:
🌲 Objective: Develop a hero image for an emailer that will raise awareness about a new marketing campaign aimed at encouraging reforesting lost tree habitats.
🏞️ Target Audience: Environmentally conscious individuals, nature enthusiasts, and potential donors.
✍️ Tone and Style: The content should evoke a strong emotional connection to nature, using compelling imagery and heartfelt messaging. It should be inspiring, hopeful, and informative, with a focus on the urgency to replant trees.
Key Elements to Include:
High-quality images of endangered/lost habitats and wildlife.
Powerful messaging that conveys the importance of tree conservation.
Branding elements such as a logo and campaign hashtag.
Let’s begin!
Step 1: Before you generate, ideate
Do a little research first.
Mainstream AI text-to-image generation has only been around for a couple of years, yet I always find someone has generated a reference image that stimulates my visual ideas.
I am an early adopter of Midjourney AI, my generative AI tool of choice. However, there are many text-to-image generators available based on budget and desired features.
I like to save generation time by starting with Midjourney’s searchable public gallery on their alpha website. The gallery is a good place to get style and content ideas before diving into personalized image generation.
Other ways to uncover ideas before wasting generation credits is to peruse the surplus of prompt creation tools available online. Here are some links to helpful tools:
💡 To address my brief, I landed on an idea to generate a composite image showing the contrast between a thriving forest and one devastated by deforestation. One side would be tree-filled and full of life and the other side deforested, barren, and dry.
Step 2: Generate using your idea
With the prework completed, I was ready to start generating images in Midjourney for my custom hero image. I started with the prompt.
“A composite image showing the contrast between a thriving forest and one devastated by deforestation.”
Examples of some of the resulting images are below.
The images I got were a promising start, but the opposing sides didn’t contrast enough or have much emotional impact.
I decided to add “child” into the prompt because children, like trees, are our future. I also remember climbing a lot of trees when I was little. The next images are the results of tweaking the prompt.
I preferred these images better, and they inspired the next step.
I decided to make my own custom composite collage. I would combine different AI generated components to create a custom split view scene like the reference above.
Step 3: Making a custom composite image with AI generated images
My initial image research and image generation saved me a lot of time so that I could get to the step I enjoy, making a visual story.
Instead of prompting the Midjourney bot to make one composite image. I focused on generating an image for each half.
Here, I implemented Midjourney’s style reference feature. By adding the command –sref random after a text prompt, the Midjourney bot randomly applies a unique style to the image it generates. After the image is generated, — sref random converts into a –sref number (i.e. — sref 4270736322). The –sref number can be added to future prompts to maintain the style.
Like prompt commands, many sites offer curated examples of Midjourney style references. One can copy and paste desired style reference numbers to apply a consistent style across multiple image generations.
For one side of my composite, I changed my prompt to include a child sitting under a lush tree. I added both a nature-stylized style reference code and a photorealistic style reference code I found in a reference library.
Below are some of the output images. I highlighted the image I chose to use with a red square.
For the other side of my composite, I changed my prompt to include a child sitting next to a dead tree in a deforested landscape. I applied the same style reference codes.
Below are some of the output images. I highlighted the image I chose to use with a red square.
With my main reference images generated, I was ready to revisit the brief and weave my hero image together into a story.
Step 4: Custom storytelling through photo editing
Photo editing software tools are great to resize, touchup, or craft custom image stories with AI generated images (or stock images).
I like to use both Adobe Photoshop and Illustrator because I have already spent a lot of time and money learning how to use them for my work. Many folks prefer to use the Canva platform, which is a great alternative.
Now that I had my selected AI images, I decided to revisit the specifics of my brief and the dual image reference I liked from Step 2.
✍️Tone and Style: The content should evoke a strong emotional connection to nature, using compelling imagery and heartfelt messaging. It should be inspiring, hopeful, and informative, with a focus on the urgency to replant trees.
Key Elements to Include:High-quality images of endangered/lost habitats and wildlife. Powerful messaging that conveys the importance of tree conservation. Branding elements such as a logo and campaign hashtag.
I started in Photoshop. I transformed each image to reflect so the young boys would have their backs to each other to mimic my reference image. I did a little touching up and color editing to match the boys’ shirts.
Then, I added text to convey a message about the two disparate sides. Below is the first draft.
Next, I started iterating around the message by thinking about the feelings the boys represented. The boy on the left looks hopeful. The boy on the right appears in despair. This led me to think about dreams and how our environment affects our hopes and dreams.
Below is the next iteration that includes a display title “Dreaming.” To address all the key elements of the brief, I included a bird, my logo, and a campaign hashtag.
In the last iteration, I decided the composite needed one more element to tie the message together to imply the state of dreaming. I added a misty haze in the center to suggest a dreamy state.
I want the audience to see that the photorealistic elements are a work of imagination.
Below is the final image.
Final thoughts
I hope this example of using AI-generated images to create a custom hero image for an email campaign inspires more purpose-driven use for your own AI image generations.
Because AI can generate images so easily, I like to complicate things by layering in my human touch.😊
Things to remember:
Before you generate, ideate
Find inspiration from others online
Delete unwanted images from your AI platform to save server space
Make things your own
Finally, let me know what you think of the final hero image?
If you like this article, you might enjoy this one too! Find your own style using Midjourney's blend mode.
Julia Fletcher is founder of JEFS Storytelling Arts, a graphic design studio, where she uses her unique research skills and artistic talents to create custom visual stories that help clients’ increase engagement and promote the education of their audience.
Comentarios