Closer #007

Not to ruffle someone's feathers but a little birdie told me to continue this newsletter until the pecking order between us, beautifully flawed and inherently creative humans, and AI, the ever logical mimicker of life, has been determined. Who comes out on top this week?
The image shared with ChatGPT as input.

Welcome to Closer,

a weekly newsletter where I take one of my photos, ask AI to describe it, then see what kind of image it can generate from its own description—an ongoing experiment in the capabilities of AI, and what that means for humans. This is edition #007.

It's February, 2022, and I am strolling along the Dutch coastline. I'm feeling a little lost and I have no real clue how to remedy that except for simply keeping it moving, hoping for a stroke of genius. I didn't find genius, but I did find this abandoned structure at the same time a flock of birds flew by in perfect formation. And that was close enough.

The order in which we operate: human experience → photograph → AI description → synthetic image. Prompts are in the grey callout boxes, the AI description is cursive.

LLMs used

GPT-4-turbo for image interpretation. DALL·E 3 for image generation (more simply said: I just prompt away in the default free ChatGPT interface and hope for the best.)

The input

✍️
Please describe the attached image as well as you can. Don't skimp out on details and make sure you will be able to understand what's in the image when you read the description. You need to be able to form a mental picture of the image, without actually seeing it.
The image shared with ChatGPT as input.
The image shared with ChatGPT as input.

The output

The image captures a quiet, melancholic coastal scene. The composition is divided into three main visual zones: the expansive sky, the distant sea, and the sandy foreground with a solitary structure.

Foreground and Structure:

In the lower third of the image, there is a white, weathered, rectangular structure that looks like a small, standalone hut or possibly an old shipping container repurposed for temporary use. It rests directly on the sand with no visible foundation, slightly tilted, suggesting it’s either been abandoned or temporarily placed. The structure has a flat black roof with at least one small metal pipe protruding from it. Its walls are plain white but worn, with a panel of raw wood visible near the base on the right side, likely covering some damage. The structure is framed by tufts of long dune grass growing along the edges of a sand dune that gently slopes toward the sea.

Beach and Dunes:

The sandy beach is pale and flat, showing tire tracks or compressed lines running along it, indicating recent human or vehicle presence. Behind the structure, dune grass clumps provide natural texture and contrast. A wooden fence is faintly visible within the dunes, acting as a subtle border between the beach and inland vegetation.

Ocean:

Behind the dune grass, you can see the choppy, grey-blue sea. The waves are gentle but consistent, visible mainly as white froth on the horizon, suggesting a slightly windy but not stormy day.

Sky and Birds:

The top two-thirds of the image is dominated by a dramatic sky filled with soft, overcast clouds in shades of grey and white, with patches of blue peeking through. Scattered across the sky are numerous birds in mid-flight, mostly concentrated in the upper left and middle sections. They appear to be seagulls or similar seabirds, flying in loose formation, with wings in varying positions. Their movement adds a sense of lightness and contrast to the stillness of the scene below.

Atmosphere:

The overall mood is serene but lonely — perhaps even slightly desolate. The muted colors, weathered textures, and presence of birds suggest a transitional season, possibly late winter or early spring. The photo has a natural, documentary feel — quiet and observational.

This is an image about solitude, time, and place — an empty beach under a soft sky, with a hint of human trace and the endless movement of nature above.

✍️
Generate an image with an aspect ratio of 2:3 based on this description: {the description previously generated}
The image generated by ChatGPT based on its description of my image alone.
The image generated by ChatGPT based on its description of my image alone.

Impressions

We're back with ChatGPT after our short stint with Midjourney, which means we're also back to essays for descriptions. I don't really like how long these descriptions have gotten. I didn't ask for it and the first couple of editions of Closer did not suffer from this. Should I set a word limit?

Anyway, the overall description is super accurate again. Much better than what Midjourney produced out of the box (as stated last week, I have a feeling Midjourney is intended to be used as a more professional tool where you are able to adjust the outcome according to your needs, using all the different options and parameters you can use to influence your output.)

One thing I particularly enjoy about this week's output is that final sentence "This is an image about solitude, time, and place — an empty beach under a soft sky, with a hint of human trace and the endless movement of nature above." Very poetic. Makes you wonder what source material it uses to come up with things like this.

The image though? Hmmm, I'm not so sure. While the subject and composition is generally correct, I think the exposure is far too dark and I can't find a real reason it decided to do that. That darkness is not really in the description, unless you count the overal moodiness it hints at of course.

The structure looks a little different from the original but there were no real parameters given in the description for how that thing actually looks so I'm not surprised here. I wish I came across a structure as nice as the AI-gen one. Other than that, I enjoy the distribution of the birds in the sky from my original image much, much more. It has a balance to it. That diagonal line cutting across the image, leading your eye through the frame is one of this image's greatest strenghts. That's completely lost in the AI output. That's probably for the best.

Are we still keeping count? Because I think the win goes to, and let me quote something somebody on my day job said last week, the "human generated content." Yes, that's a real sentence that was spoken. But then we also laughed about how ridiculous that sounds, thankfully.

Anything else that caught your eye? Feel free to send a reply. Please also let me know if you have questions, suggestions, or remarks about my approach. This is just as much a journey for me as it is for you.

See you next week.

Mitch

Subscribe to the monthly newsletter

Just you, me, and some occasional notes from the field. No spam.

Join the conversation