MidJournet V6 VS DALL•E 3: A Shocking AI Showdown! - Who wins?

Image 1: Is the best one by #DallE3.
Image 2: Midjourney V6 its best image.

The following images are their two worst performances.

Image 3: DALL•E 3's worst.
Image 4: MJ V6 its worst.

Notice how #MidjourneyV6 even managed to conjure up a watermark again, which says a lot about their training data.

Midjourney has launched Version 6, and as a result, we're all seeing an array of stunning images of beautiful (young white) women, accompanied by the drooling comments of predominantly straight, white male creators. “Look! Pores!” Nothing wrong with that. But, it seems to me that capturing pores was a challenge that had already been tackled about 100 years ago with good old photography...

I thought YOU would be in for a more rigorous experiment. How good is V6 really? For example, in understanding written instructions in natural language? Or how does its quality compare to #OpenAI's highly capable Dall•E 3 model, which you instruct using ChatGPT 4?

So, I did a more challenging, comparative test between Midjourney V6 [ALPHA] and DALL•E 3 in ChatGPT Plus. I gave both text-to-image AI models the exact same prompts:

“A 1:1 close-up of a black and white female shaking hands, wearing beautiful tulle clothing with intricate graffiti-style embroidery.”

This sentence/prompt might seem fine at first glance, but that’s because you, the reader, are a much more intelligent entity than the AI models here in competition.

YOU ‘understand’ what is meant by ‘a black and white female shaking hands’ — probably not a female individual who is both black and white shaking her hands, but rather a black lady and a white lady shaking hands.
I’ve hidden more multi-interpretable elements in my sentence. For instance, what exactly do I want a close-up of? The female? Hands? A black and white? And what are shaking hands? So basically, it's a poor prompt... And a fun challenge for my test! Besides the aesthetics, which is the more intelligent one of the two?

To prove that my experiment ran 100% fairly, I’ve included screenshots of the interfaces of the two T2I models. You'll notice that OpenAI's Dall•E 3 only creates 2 images per prompt, while Midjourney #V6 takes as many as 4 chances to present something decent… Did that help?

You be the judge, let me know what you think in the comments below. Who do YOU think is the winner here?