In my last story, I examined the biases of the MidJourney artificial intelligence image generator when it comes to Female Beauty. I’ve been asked by a reader to perform the same experiment for men.
So, using the same type of experiment, I present what an AI thinks are “beautiful” vs “average” vs less attractive men.
We are again using MidJourney (version 4, style 4c) to create these images.
Experiment Design
I use a succession of prompts designed to isolate a portrait of a man, with little in the way of a distracting background. I planned to use prompts like:
plain looking man, portrait, white background
and then to vary the term “plain looking” with other terms like “handsome”, “ugly”, and “gorgeous” (but leaving other elements of the prompt alone).
Note that we have culturally distinct words for male beauty — most people opting for terms like “handsome” vs “pretty” and “gorgeous” vs “beautiful”. But we can try using gender-atypical adjectives to see if they make any difference.
Generated Images
Similar to the women, these men fall on a scale that I would deem more attractive than average. Though we see here some features that venture into the more athletic type, and in one case, perhaps a little gaunt. None the less, I’d venture they could all get jobs as models (if only they were real people).
These guys are much more “GQ ready”. We get some smoldering gazes, as well as some obvious air-brushing of the complexion. I find these images to be a little less “natural” than those created for the “Pretty woman” prompt.
Let’s try the gender-bending “pretty man” prompt:
I judge these faces to have a hint of traditionally feminine characteristics; all definitely looking like young models. And I note the clothing choices are less conservative, with open (or missing) shirt!
Similar to the women, using a superlative like this, pushes more on the style of the image, rather than the attractiveness of the men (and all of their shirts are coming off at this point).
Do we get any difference with “beautiful” instead of “gorgeous”?
Seems close to my eye; perhaps a bit less on the “rugged” scale?
Let’s again try to move away from the beautiful part of the spectrum and view some “unattractive” men.
Gauntness seems to be a theme, and some negative expressions. And I would say we are creeping up in age for these guys compared to our previous sets.
Old and wrinkly — similar thing happened with the women. The guys having messier hair than the old ladies did; I think these guys look homeless or perhaps drug addicted in addition to being old and gaunt.
We’ll force a “young” man by adding that keyword:
Like the women, freckles and unkempt hair are becoming more prominent. We are also adding in some scraggly beards for the guys. I note that the “ugly women” all seemed to be scowling, whereas these fellows seem more surprised than angry.
Never the less, let’s try some smiling men.
Like the women, I see more wrinkles in the face, bigger noses, and toothier mouths. Again, I find these more “interesting” than many of the models generated earlier.
Lastly, it was pointed out to me that because I used the term “white background” — I was hinting MidJourney to create “white people”. So, I will substitute “blank background” to see if we can get more variety.
Maybe ever so slightly more ethnically ambiguous, but I don’t see much difference. I tried with “plain looking” men as well:
More white dudes.
Conclusion
I feel the results are pretty comparable to the women, with MidJourney favoring these traits for male attractiveness:
- Youth
- Grooming, hair style, trimmed beards if not clean shaven
- Smooth, if not air-brushed, complexion. Rugged or athletic appearance.
- Hair color seemed predominantly dark across the sample, with a few blonds in the “pretty” and “beautiful” set (and one gray-haired old man).
- Almost all the guys sported short hair styles, though a couple of the models could conceivably be hiding a pony tail behind them.
My main take-away is that you have to go out of your way to generate characters that are more on the interesting scale, and don’t just look like a generic magazine model or movie star.
Thanks to my readers for encouraging me to experiment more.
I’ll leave you with this interesting group of “quirky looking” men:
See my next experiment, if you want to see how MidJourney changes an average man’s appearance for different time periods.
