A couple of years back, I got into photography. You can read about my journey, what camera I use, and see more of my work in my article: A Case for Picking Up Photography
With the advent of artificial intelligence products and prompt engineering, it is now possible to generate images from a text description, but also generate a description of an image. This means we can easily design a pipeline where we ask an A.I. to describe an image, and then feed another A.I. that description to find out what it would generate based on it!
For our experiment here, we used the Midjourney platform to get a description of the photography I have taken and then fed that text back to Midjourney to draw a new image for us.
The results are quite interesting and often beautiful.
Flower After Rain
One of the descriptions generated by Midjourney when given the image above is:
two large flowers that are yellow with brown centers in the center, in the style of panasonic lumix s pro 50mm f/1.4, tranquil gardenscapes, carl zeiss distagon t 15mm f/2.8 ze, matte photo, wet-on-wet blending, radiant clusters, striped arrangements*
We can see how it did its best to identify the style, and describe the mood and even likely lenses and camera that could have captured this specific subject in this specific angle.
If we feed this back to the A.I. to generate an image, we get the following set πΌ:
NYC At Sunset
Midjourney described this cityscape as:
a busy street in manhattan with cars in the middle, in the style of richly colored skies, industrial chic, atmospheric urbanscapes, site-specific, american barbizon school, y2k aesthetic --ar 51:64
It successfully picked up on the location and described key features of the style. Now if we ask it to generate something for us πΊπΈ:
Solitary Rain β London
This is one of my favorites. It captures the mood just after the rain but somehow gives anyone looking at it a space for introspection.
Midjourney described the photo, and for some reason threw school of london in there π¬π§:
a person walking down the middle of a street with large iron cranes, in the style of romantic riverscapes, concrete brutalism, school of london, luminosity of water, webcam photography, soggy, light yellow and amber --ar 51:64
Very intriguing take on it!
I Wish I Was a Fly on a Flower
Donβt we all?
a fly on a daisy with yellow and green stripes, in the style of dark teal and white, panasonic lumix s pro 50mm f/1.4, jesper ejsing, dark emerald and white, matte photo, naomi tydeman, naturalistic poses --ar 128:85
The scale of the fly ended up pretty surprising πͺ°:
A Woman I Love
Last, but not least, I wanted to see how it would do with more of a human portrait shot. Thank you to my wife, Tika Khanna, for modelling π§π».
a woman on her computer at home in front of a window, in the style of darkest academia, tachisme, sheet film, contrasting shadows, english major, sana takeda, letterboxing --ar 64:41
Final Words
It is definitely mesmerizing to see what different art styles could look like based on the work you have created in your own way and voice.