Testing 3 different AI platforms to generate images

Testing 3 different AI platforms to generate images

Everything is simple until it’s not.

I’ve been using generative AI for images for a while now. I consider myself okay at prompts, so when I needed an image for an event we’re putting on in NYC for Fintech Founders on April 17th, I immediately thought about creating an image using generative AI instead of spending hours searching for the right one.

I couldn’t judge Gemini, as it’s currently shut down its ability to generate people.

Also, I didn’t use the same prompt the whole way through. When you see the images, you’ll see that there was no point as only one image generator did a good enough job to be usable. 

For those of you with no patience, here’s the TL; DR

 ChatGPT (DALL-E)META Gen AI inside WhatsAppMidjourney

Photo

realistic

Not even close.Kind of?Better, but not perfect.

Men and women

without specifying

Yup. I liked how one picture included a black woman with natural hair.NopeYes.

Looked like

the prompt

Oh dear.Kind of?Yes.

Good enough

to use

No. Absolutely not.NopeYes.

I expected this to take me, max 30 minutes. Maybe an hour.

Buckle up, this is going to be a bumpy ride.

Testing DALL-E (ChatGPT 4)

First stop was DALL-E. I sign in and get the following message:

Using generative AI for marketing imagery

Dall-E being unintuitive.

Since that’s small, let me make it easier for you: “DALL-E 3 is now available to all ChatGPT Plus and Enterprise users. Only DALL-E 2 is accessible on the current page you are visiting on.”
So, I move to ChatGPT, because I pay for Plus and I have seen DALL-E 3 images online, it’s supposed to be way better.
Um, either I don’t think I have access to DALL-E 3, or it’s not way better.
Here’s the prompt: You’re a senior photographer at WSJ and you’re taking pictures of a Fintech marketing workshop that is in a boardroom on the 6th floor of the world trade center. There are 25 people in the room. Create a photo that represents that scene.

Fintech workshop image

This isn’t what I was looking for. The prompt is on me.

I changed the prompt: create a photo of men and women wearing suits at a networking event using DALL-E 3.

Men and women networking in a room - marketing image

Oh dear. Somehow this is even worse.

And because I try things at least 3 times before declaring it a failure, I tried again! This time the prompt is: Create a realistic photo of executives in a room talking to each other at a networking event.

marketing imagery using generative AI

DALL-E’s version of photorealistic. Yeah, no. Also, look at her feet.

Next up – Meta’s image generator inside WhatsApp.

Testing META’s generative AI inside WhatsApp

Prompt: Photo of men and women networking. They are all wearing suits.

Using Meta generative AI for marketing imagery

This was unexpected.

Trying again with a new prompt: Create a photo of people wearing suits talking to each other at an event.

Apparently, I need to specify women?

Final try: Men and women at a networking event in NYC talking about finance. The interior has windows on all sides and is in a skyscraper. The people are CEOs of Fintech startups.

I specified women. Still, not quite what I was after. And the people are too pixelated.

Testing Midjourney

Midjourney (side note – if you like cars and fashion, check out Car.Obscura on Instagram)

I know Midjourney does photorealistic better than DALL-E, so that’s the next stop.
 
Prompt: networking event for people in finance. Photo of men and women in suits standing around talking to each other. Realistic.
 
With Midjourney you get 4 responses:

The guy holding the paper almost looks like a waiter. Also, his right hand has 5 fingers and no thumb?

What is her blouse doing? This one is better, but I wanted men and women’s face. Realistically our audience will be 90% male.

This is the one we started with. Don’t love it. No woman in Fintech would let her hair get this sloppy.

This is the one we ended with. Again, what is with her hair?!

Midjourney won this round hands down.

I went back to Midjourney and tried One Last Time.

This time I googled the top professional camera and used that for the prompt:
photo taken with a Canon EOS R5 C RF24-105mm F2.8 L IS USM Z Lens of a room of professionals networking. Everyone is well put together and wearing suits. Men and women. Multiple races. They are in a boardroom on the 6th floor of the World Trade Center

Her hair looks the best so far. I don’t like that the center of the photo is the back of someone’s head.

He looks like he was a boxer in his past life?

I’m going with this one.

What do you think? Did I pick the right one?

If you are a Fintech Founder in, or near, NYC – come to our workshop: Fintech Marketing Masterclass: Tell your story so a bank will buy.

1024 1024 Hunter & Bard
Start Typing