Unless you’ve been living under a rock, you cannot have missed the groundswell of stories about ChatGPT and its potential to change how we do just about everything in our field of public relations.
Much has been written about ChatGPT’s ability to help with the writing, outlining and generating ideas for PR pros. That’s because ChatGPT is the conversational side of OpenAI’s suite of products.
But there is another, less publicized, tool in the OpenAI suite that you should get to know immediately.
DALL·E 2 (pronounced “dolly two”) is to graphics what ChatGPT is to words. I am not exaggerating when I say it could change your life. I made a presentation on it to our agency last week, and my colleagues were amazed.
Why?
Simply put, DALL·E 2 can produce in seconds what a graphic designer may need many minutes or even hours to produce.
And here’s the best part – because it is generated by artificial intelligence, there are fewer copyright and licensing issues. Users are encouraged, but not required, to disclose images that are AI generated. It is also forbidden to tell people the images are human generated or are photographs of a real event. Yes, the images can be that good.
There are three immediate PR use-cases for DALL·E 2 graphics:
- Use in blog posts that require some graphics to break up long running text
- As a tool to generate social media graphics
- Generation of conceptual images that can be given to a graphic artist for further development
There are some guardrails that OpenAI has placed on DALL·E 2 images. The tool will not generate images that are violent, hate, or porn. It will also not use images of real individuals, including public figures. This cannot be said for all image generation tools.
To use DALL·E 2, sign up at https://openai.com/dall-e-2/. Then give it a command. Free accounts have access to 15 credits per month. To generate more images, additional credits can be purchased that work out to about 13 cents per image.
At our Lunch ‘n’ Learn last week, our team played around with DALL·E to learn what it could do. Here are the commands we gave it, followed by the images it returned.
Curious to see if we could generate an image to accompany a Super Bowl blog post we had written, we gave DALL·E this command:
“Two NFL fans, one male and one female,
wearing green uniforms standing in front of the Philadelphia skyline”
Here’s what came back, in mere seconds.
DALL·E will generate four images for each request. Each request counts as one credit. It will not generate photos that contain copyrights or trademarked images, so specifying uniforms of a color is about as close as you can get to a trademarked jersey.
Remember the people in these images are composites. Pure fabrications by the AI tool. They are not models; any likeness to a real person is purely coincidental.
For our second command, we got a little crazier.
“Famous singer drinking pickle juice on a beach”
It is not unusual for the AI tool to not comply with the request exactly. In this request, the images are of people eating pickles, not drinking the pickle juice. It happens.
Get the idea?
Go to the DALL·E 2 web site, create an account, and play around. It’s a lot of fun! And it could seriously speed up your turnaround time for PR deliverables that need a graphic.
# # #