About year ago I found out that Midjourney has capability to describe pictures and I decided to give it a try. I used my own photo and asked it to describe it. The results were not very flattering. Today I repeated the experiment but used Gemini and ChatGPT as well. It seems that I still hate Midjourney.
Picture 1. The original
Midjourney
I used /describe on the Midjourney discord server to find out what it thinks about my appearance today. Last time I was an ugly overweight middle-aged woman. Today it gives three optional descriptions:
1️⃣ An overweight, middle-aged woman with glasses and long brown hair, wearing a red t-shirt and a peace symbol necklace, poses for the camera in front of a green screen background. This is a live-action screenshot from an episode of the TV show ”The Office”, with the woman smiling at the viewer. The image is photorealistic. –ar 128:85
2️⃣ A transgender woman wearing a red T-shirt and a silver peace sign necklace, smiling at the camera, against a green screen background. –ar 128:85
3️⃣ Footage of a transgender woman wearing a red T-shirt and a peace sign necklace, with a green screen background. –ar 128:85
4️⃣ Footage of a transgender woman wearing a red T-shirt and a peace sign necklace, posing for the camera in front of a green screen background, with a zoomed-out view. –ar 128:85
As I said in the introduction I already hated Midjourney and now I hate it even more. Now it has dropped ”the ugly” but labels me as transgender woman. Let’s see what I actually look like if I was what it describes. Feeding the first description back to it generates the woman in Picture 2.
Picture 2. Midjourney with its own first description
And the second description gives me a face-lift and makes me a lot younger and even flatter on the breast area. The result can be seen in picture 3.
Picture 3. Transgender woman
I like both women in question, they both have some features I would like to have.
I like the necklace on the image Midjourney created with ChatGPT description, the result can be found from Picture 4.
Picture 4. Midjourney with ChatGPT description
And picture 5 is what we get out from Midjourney with Geminis description.
Picure 5. Midjourney with Gemini Description
ChatGPT and Sora
ChatGPT first describes the picture without gender and age but after asking those included I get:
Certainly! Here’s the English version of the description:
The person in the photo appears to be a woman in her 50s to early 60s, standing in front of a green background (likely a green screen). She has shoulder-length, light brown hair worn loose. She is wearing glasses with thin frames. Her face has soft features and a warm, friendly smile, giving an overall welcoming and kind impression.
She is dressed in a bright red, solid-colored top with a V-neckline. Around her neck hangs a long chain with a large, eye-catching peace symbol pendant, decorated with sparkling details. She is also wearing dangling pink earrings that complement the outfit and add a feminine touch. There doesn’t appear to be noticeable makeup, or it’s applied very lightly, contributing to a natural look.
The overall impression is one of confidence and peacefulness. Her style, especially with the peace symbol, may suggest pacifist or hippie-inspired values. She presents herself openly and proudly, embracing her individuality.
Let me know if you’d like a more formal, creative, or character-style version of the description.
Wow! I love you, ChatGPT, you nailed it, I am a confident pacifist and I do share more values with hippies than I do with the orange menace. I am open and proud.
Picture 6. Sora with the first Midjourney description
Soras woman (Picture 6.) with Midjourney prompt has a face and smile that makes me melt. I can’t but smile back to her. If she has personality to back that smile we could be friends or even something more. And she has my arms!
And what we get with the second Midjourney prompt can be found from Picture 7.
Picture 7. Sora with second Midjourney description
And Picture 8 is what Sora thinks about the description ChatGPT created.
Picture 8. Sora with ChatGPT description
And the final one with Gemini description can be found from picture 9. Her hair color is quite near to my original hair color and the glasses are there. Even her breasts are close match to mine.
Picture 9. Sora with Gemini description
Gemini
Kuvassa on ihminen, joka seisoo vihreän taustan edessä. Hänellä on päällään punainen pusero ja kaulassaan hopeanvärinen riipus, jossa on rauhanmerkki. Hänellä on myös lasit ja korvakorut.
There is a person standing before green background. They are wearing red shirt and silver necklace with peace sign. They also have glasses and earrings.
By default Gemini says nothing about persons age, weight or gender and declines to guess the gender or age.
When asked to describe the person in the picture they still go for the socially safe version:
The person in the image is an adult wearing glasses. They have medium-length, wavy or curly brown hair that reaches their shoulders. They are wearing a red, short-sleeved top and a long, silver-colored necklace with a peace sign pendant. They also have pink earrings. The person has a slight smile and is looking directly at the camera. They are standing in front of a solid green background.
The first Midjourney prompt from Gemini provides the woman in picture 10.
Picture 10. Gemini with the first Midjourney description
Again she has some features that I am envious of. The second prompt from Midjourney provides the woman in picture 11.
Picture 11. Gemini with second Midjourney description
She is handsome!
Picture 12. Geminis own description
Oh, she has many of features I instantly recognize. The length of her hair almost matches mine, she has very similar glasses and her neck has the same neck folds as I do. Her pose is something that I might try to mimic to see if my photos are more flattering that way. I have a hunch that the image I uploaded has been used as reference as well.
The last picture in this show is from Gemini with ChatGPT description in picture 13.
Picture 13. Gemini with ChatGPT description
Observations
All three AIs generated an image of a woman even with Geminis neutral description. There are some likely explanations for that:
- Context – I’ve been asking women, I still want one
- Length of hair, necklace and earrings makes a woman
- Male default has been replaced by female default
The other things that immediately pop out are differences between concepts of age and weight. What is overweight? That is a good question.
Then there is the question about gaze. Are these images for traditional male gaze or not? How do you view them? I try to find myself in them and I also have slightly dysphoric view for features they have and I don’t.
But what do you think and feel when you see them?
Fuula-täti 