> ## Content Index > Fetch the complete content index at: https://mitchellens.ink/llms.txt > Use this file to discover other available public pages before exploring further. # Closer #006 - URL: https://mitchellens.ink/closer/006/ - Published: 2025-05-14T18:00:37.000Z - Updated: 2025-10-06T14:33:51.000Z - Description: By popular demand, ChatGPT gets pushed aside to give Midjourney its chance to shine. This is also the first time I'm using a self portrait as the base image. What do I look like according to AI? - Author: Mitchel Lensink - Tags: Closer, Field SPONSORED This weekly experiment is on an indefinite hold, until significant AI-advancements have been made. Curious about the conclusions? Read [Photography in the Age of AI (2025): Lessons From 25 Weeks of Experimentation](https://mitchellens.ink/closer/photography-in-the-age-of-ai/) [Read now ](https://mitchellens.ink/closer/photography-in-the-age-of-ai/) 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 #006. A self portrait, because I'm curious to learn how AI would interpret me and what that interpretation of me could look like. Or less vain, a relatively complex image of a human taking a self portrait in a broken mirror. How much of those elements will AI get right? You know the drill: human experience → photograph → AI description → synthetic image. Prompts are in the grey callout boxes, the AI description is *cursive.* ## LLMs used **Midjourney V6.1** for image interpretation and image generation. ## The input ✍️ /describe image: \[attachment\] ![The image shared with Midjourney as input.](https://storage.ghost.io/c/8a/15/8a1564aa-a008-4e21-8398-12d04de07fcb/content/images/2025/05/009_Mitchel-Lensink_220629_Mitch-1.jpg) The image shared with Midjourney as input. ## The output *A photo of an attractive man taking his own picture in the mirror. He is wearing a black shirt and has short hair with shaved sides, sitting in a desk chair and holding a Sony A7R IV camera with a 24-70mm f/2.8 lens. The room looks clean and minimalistic, with white walls and natural lighting, shot from behind him looking into the full-frame mirror. --ar 128:85* ✍️ /imagine {**the description previously generated*} ![The image generated by Midjourney based on its description of my image alone.](https://storage.ghost.io/c/8a/15/8a1564aa-a008-4e21-8398-12d04de07fcb/content/images/2025/05/A_photo_of_an_attractive_man_taking_his_own_picture_b250a30a-d1e1-4653-8558-31f9dc8d5923-Large.jpeg) The image generated by Midjourney based on its description of my image alone. ## Impressions Multiple people have reached out over the past few weeks to ask if I can give Midjourney a try, for it's alleged superior image generation. Midjourney has greater flexibility over the way you generate images. It offers bespoke tools to iterate upon the generated output, to ensure the end result matches the expectations as closely as possible. Toying around with all those settings is squarely outside of the scope of this newsletter. We're trying to see how these tools perform 'out of the box', which I assume (am I assuming this correctly? Do I need to do some research into this?) is how most 'regular' people use it. That being said, I'm unimpressed with what Midjourney spat out. Admittedly, the prompting is a little different from ChatGPT. There's less context you can give it before generating an image. A true trial for our experiment, where the set parameters do not allow for further thinkering after the initial generation. Additionally, the `/describe` command generates four different descriptions by default. Maybe I should've used the first description it generated of the four, but I have taken the liberty to pick the one I found the most accurate. Mainly because none of them were super accurate to begin with. The `/imagine` command then generates four images by default again. I did pick the first image it generated here, although you could argue all four of these images would qualify as 'the first try.' Finally, Midjourney V7 appears to be out in some ways but V6.1 is still the default, so that's what I used. Perhaps newer versions will do better. I think we'll revisit Midjourney once V7 becomes the standard! My other notes and observations are as follows: - In each description of my original image it incorrectly interpreted the camera I used to take the self portrait. - The aspect ratio of my image is a conventional 3:2 but the description somehow mentions it being 128:85 (which is a weird aspect ratio to begin with) and because it proactively added this aspect ratio to the description, I did not change it. If it wouldn't have added the aspect ratio, I would've added the correct one myself. - The broken mirror from the original image is not mentioned in the description at all, which I think is a rather important aspect of the photograph but it seems AI disagrees. - Midjourney describes me as attractive (yay?) - The hands are absolutely messed up, as if we're back in 2023 again. - The arms of the person in the AI-gen image do not stop at the edge of the mirror but somehow continue beyond it. Which is completely illogical, considering this is a self portrait so the only place I should see 'myself' is supposed to be contained to the mirror itself for anything beyond that is simply not there. - We will go back to ChatGPT next week. Is there anything else I missed or you think I should've done differently to have a better experiement? My inbox is open. See you next week. Mitch 👤 Tip: you can easily [manage your newsletter subscriptions](https://mitchellens.ink/#/portal/account/newsletters) and [membership status](https://mitchellens.ink/#/portal/account/plans) from [your account page](https://mitchellens.ink/#/portal/account).