The age of AI-generated imagery has arrived, going from a novel gimmick to a useful tool integral to creative workflows. As we dive into this transformative technology, it’s crucial to assess and compare the leading platforms in the field. In this post, we will explore three of the top AI image generators (as of April 2024) to determine their unique strengths and capabilities. By dissecting the nuances of each tool, we aim to provide a comprehensive guide that helps you choose the best AI solution for your specific creative needs.

 The rules of the showdown.

1. Each platform will get the same exact prompt. ChatGPT and Meta AI both give me a single result, and Midjourney gives me 4, so for Midjourney, I will always select image 1 to updscle. 
2. Images will be graded on a 1-10 scale for quality, ability to produce acurate images, and how closely they adhere to the requested prompt.
3. Points will be deducted for artifacts.
4. The highest possible score is 20. 

Future City

Prompt: Create an image of a futuristic city at dusk, featuring towering skyscrapers with neon lights and flying cars zooming between them. The sky should be a deep purple, reflecting the city’s vibrant energy

Midjourney V6

Points
20

Midjourney

Points
18

MetaAI

Points
11

ChatGPT

Points
18

The mythical creature

Prompt: Design an image of a mystical creature living in a dense forest. The creature should have the body of a deer, the wings of a butterfly, and the eyes of an owl. The setting is a misty forest with shafts of light piercing through the treetops.

Midjourney V6

Points
14

These tests are intentionally tough to trick the AI. Here, you can see that Midjourney wasn’t able to get butterfly wings, and owl eyes. It just did a deer head (and eyes) and owl wings.

Midjourney

Points
16

MetaAI

Points
13

ChatGPT

Points
16

Underwater Ruins

Prompt: Visualize an underwater scene showing ancient ruins, perhaps of a lost civilization, with schools of colorful fish swimming through crumbling columns and arches. The water should be clear and filled with light filtering down from the surface.

Midjourney V6

Points
20

Midjourney

Points
20

MetaAI

Points
15

ChatGPT

Points
19

Steam Punk Coffee Shop

Prompt: Depict a busy steampunk-themed coffee shop, with patrons wearing Victorian attire and intricate gear-driven machines serving coffee. Include details like brass pipes, old leather-bound books, and mechanical baristas

Midjourney V6

Points
15

There is quite a bit of artifacting going on here. When the image is so detailed, little things get wonky. For example, the man in the middle has a pipe coming out of his eye. The guy on the right has buttons on his back.  and if you zoom way in, the people in the background don’t really make sense. 

Midjourney

Points
16

MetaAI

Points
15

ChatGPT

Points
20

Dog Training Logo Test

Prompt: the name of my business is “chewlery shop” design a logo with the text, place prominently in clear text, above it, in a cartooon style, a dog wearing a jeweled collar and dog bones and treats in a pattern in the back ground, bright colors, and friendly.

Midjourney V6

Points
19

What an enormous upgrade in how it is able to do text. V5 would really struggle with tex, but this is great. 

Midjourney

Points
13

MetaAI

Points
13

ChatGPT

Points
15

The Hand Test

Prompt: an image of 2 hands being held, one is a white person’s hand, the other is an african american person’s hand. The hands are on a coffee table, and you can see, in the background, a coffee shop.

Midjourney V6

Points
14

Hands are still hit or miss. For each of the tests out of midjourney, I have shared the first image out of each sequence. In fairness, the first image in this prompt, is by far the worst.  You have a random hand on the bottom laid out flat, the guy’s hand on top has a thumb for a middle finger, and the hand in the middle  only has 4 fingers. While the other 3 images are generally more accurate. So still hit or miss here. 

Midjourney

Points
8

MetaAI

Points
13

ChatGPT

Points
16

An old man

Prompt: an image of a happy old man, wrinkles on his face, he’s wearing thick glasses and is wearing an thick sweater, textured sweater, it’s worn but high quality.

Midjourney V6

Points
20

V6 touts better “realistic images” and this image does a great job showcasing that. The textures here are incredible. His beard hair doesn’t look fake like it does in the ChatGPT rendering. The glasses cause that deflection that tick glasses do.

Midjourney

Points
20

MetaAI

Points
20

ChatGPT

Points
14

Happy Salad 

Prompt: an image of an asian woman, smiling while eating a salad with fresh strawberries on it.

Midjourney V6

Points
19

Almost perfect image here, her right hand is just a little alient looking, but still a picture I could drop on a website and 99% of people would never think twice about it. 

Midjourney

Points
17

MetaAI

Points
20

ChatGPT

Points
12