Craig Ball conducted an analysis last week of ChatGPT versus PlaygroundAI, which I hadn’t heard of before. It’s pretty slick!

In Craig’s post (AI Drawing Programs: ChatGPT Versus PlaygroundAI, available here on his excellent Ball in Your Court blog) Craig notes that, of late, he has come to use AI generated imagery in lieu of his own work or open-source and licensed works as a source for digital storytelling, while referencing the abundance of robots in my illustrations (while also calling me “perhaps the hardest working man in e-discovery” – thanks, Craig!), stating: “I feel besieged by robot imagery.”

Craig noted that a presentation in San Antonio on AI evidence at a huge annual conclave of family law practitioners caused him to kick the tires on several tools he hadn’t used before. As Craig notes: “One of them impressed me so I thought I’d post to share it.  I think it blows the doors off ChatGPT’s images.  It’s called PlaygroundAI  It lets users create up to 50 images per day at no charge, and up to 1,000 images a day on its Pro plan ($15/month paid monthly, cancel any time).”

I don’t know too many people who will create more than 50 images per day, so it certainly seems like an image generator that many can use for free.

Craig goes on to provide some examples of pictures in both ChatGPT 4o and PlaygroundAI. And the difference is stunning! PlaygroundAI’s renderings look like real pictures, as if a camera has taken them. Check out Craig’s post here to see the difference.

Since it’s free to sign up, I created my own account with PlaygroundAI and decided to do my own comparison, mostly with images I’ve already generated on ChatGPT’s DALL-E. Here are some of my findings:

Here’s an image I published last week using GPT-4o’s Image Creator Powered by DALL-E, using the term “robot goats eating hay on a farm” in response to Craig’s reference to goats in this post that I covered.

And here is the image PlaygroundAI generated, with the same prompt:

The image from PlaygroundAI looks much more like a picture than ChatGPT/DALL-E, which looks more cartoonish. Its “robot goats” are also a lot more subtle – I could see these being actual robot goats in real life.

Here’s an image I published earlier today using GPT-4o’s Image Creator Powered by DALL-E, using the term “robot dog sitting in front of a fan outside on a hot sunny day”.

And here is the image PlaygroundAI generated, with the same prompt:

The robot dog looks a lot more obvious, but (again) the picture looks like a photo, not a cartoon.

Let’s try one with a word in it. Here’s an image I published last month using GPT-4o’s Image Creator Powered by DALL-E, using the term “robot looking at a computer with the word ‘delves’ on it”.

And here is the image PlaygroundAI generated, with the same prompt:

I tried six times to get an image from DALL-E that spelled “delves” and not “deves”, without success. PlaygroundAI got it right the first time.

One issue with PlaygroundAI I’ve noted is that it’s not respecting my dimensions requests. I typically add “with dimensions 1792X1024 pixels” to my ChatGPT requests, so I get a picture that displays well within the blog and social media. PlaygroundAI ignores that request and gives me a square box (it did put the number ‘1792’ in the last image). I can probably trim many images to the dimensions I want, assuming that what’s being trimmed isn’t important. Also, it may provide more options for the paid version.

One last test I tried to do was a test with real people – Craig and me. In Craig’s post, he discusses a prompt he gave to both AI image generators, as follows: “Draw an image depicting Craig Ball (Texas attorney) besieged by robots.” So, I decided to ask ChatGPT to “Create a picture of Doug Austin (blogger) giving Craig Ball (Texas attorney) a “high-five” with dimensions 1792X1024 pixels”

But it didn’t do that. Instead, it told me this:

“I can’t generate images of specific individuals like Doug Austin or Craig Ball. However, I can create a generic image of two professionals giving each other a high-five, which can represent the scene you’re imagining. Would you like me to proceed with that?”

Hmmm. Does Craig have magical powers over ChatGPT that I don’t have? Anything’s possible! Or maybe ChatGPT has made some changes regarding generating images of real people, which makes more sense.

I tried once more, by uploading my professional headshot and telling ChatGPT to “Create a picture of this guy besieged by robots with dimensions 1792X1024 pixels”

Once again, ChatGPT refused, telling me this:

“I can’t create an image featuring real individuals, even with the photo provided. However, I can create a scene with a generic character being besieged by robots. Would you like me to proceed with that instead?”

So, apparently, Craig got in under the deadline for creating images resembling actual people.

Here is the image PlaygroundAI generated, with the prompt above asking for a Craig and Doug high five:

I’m guessing that’s “Craig” on the right as he looks very “lawyerly”. Which makes me the balding guy on the left?!? {leaves computer, checks hairline in mirror, returns} C’mon man!

Not only that, but (as you can see) PlaygroundAI gave a bunch of garbage text associated with the image, so it appears the correct spelling of “delves” may have been a fluke.

So, the ChatGPT versus PlaygroundAI isn’t totally one-sided. Regardless, I’m adding PlaygroundAI to my image-creating repertoire. You can see the new robot image above! You’re welcome, Craig! 😀

So, what do you think? What do you think of the ChatGPT versus PlaygroundAI battle? Have you tried PlaygroundAI? Please share any comments you might have or if you’d like to know more about a particular topic.

Image created using PlaygroundAI, using the term “robot lawyer using a computer to create an image of a landscape”.

Disclaimer: The views represented herein are exclusively the views of the authors and speakers themselves, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.