Musings in mirrors

Mirrors.jpg

For this sequence of generative art, I wanted to focus again on the concept of reflection, introspection and of the "mirror". I decided to focus on a golden hue as my colour palette here, inspired by the Autumn colours which have departed the locale in which I reside, instead replaced by Winter in a Spring that doesn't seem to want to get started, apart from the pollen.

I began with a prompt that returned the following:

image.png

she looks into the mirror beautiful attractive women, hourglass figure, beautiful face, golden atmosphere, mysterious, magical, golden hour, crepuscular rays, hyper realistic, unreal engine, fantasy, 8k, octane render, extremely detailed and complex, hyper-realistic, enhanced, photo render --ar 8:10

image.png

I thought that the first image in this sequence was quite striking, so decided to upscale it, but it was a no-go, losing too much detail from the photographic style of representation I was hoping for in this image.

image.png

Time for some further iterations, and the characters depicted in concepts 1 and 3 are absolutely stunning, so I decided to upscale these, even though they were not related to the concept I am pursuing.

image.png

Here we can see image four upscaled:

image.png

Again, there's a loss of detail; the work doesn't hold up to scrutiny when observed at actual size or larger than a thumbnail. This leads to the work having a certain quality of what a scholar like Walter Benjamin would call "mechanical reproduction", which is in a way, exactly what this is.

How, I wonder did the first image upscale fare?

Far, far worse. I'm removing some details from the prompt for the next iteration in the hopes that the game engine references will be removed in order to obtain a sharper image. Failing that, I will add some terms regarding photographic light and compositional cues.

image.png

/imagine prompt:a woman's reflection in a mirror; hourglass figure, beautiful face, golden atmosphere, mysterious, magical, golden hour, crepuscular rays, hyper realistic, 8k, octane render, extremely detailed and complex, hyper-realistic, enhanced, photo render --ar 8:10

image.png

This got me some interesting results (particularly number one) - but time to change the prompt to see if I can push this more into the creative direction I desire.

/imagine prompt:woman's reflection in a mirror; a photograph shot at f8, 35mm, iso 50, studio lighting, hourglass figure, beautiful face, golden atmosphere, mysterious, magical, golden hour, ray tracing, annie liebowitz, ansel adams, 8k, octane render, extremely detailed and complex, hyper-realistic, enhanced, photo render --ar 8:10

This resulted in quite the abstract interpretation, even though I felt the prompt was far more literal than anything I'd provided prior. Here are some variations before I changed it further

image.png

image.png

Number 3 here is rather striking, as is the structure of number four, but the prompt can be further refined to get a different outcome, I think:

a woman looking at a woman's reflection in a mirror; a photograph shot at f8, 35mm, iso 50, studio lighting, hourglass figure, beautiful face, golden atmosphere, mysterious, magical, golden hour, ray tracing, annie liebowitz, ansel adams, 8k, octane render, extremely detailed and complex, hyper-realistic, enhanced, photo render --ar 8:10

image.png

Again, three and four are the most striking images - but they're not quite where I want them to be. The 2nd image has a mirror present, and a woman, but there's not the connection made. I could conceivably blend image three in photoshop and make a reflection be quite compelling, and I think that may be an experiment I run if further iterations do not continue to approach my ideation.

I ran an experiment on a further version in Photoshop before having a bit of a breaththrough.

image.png

This is a very rough mock up, but plays on the mirror image concept in some more detail with a random face that looks similar to the primary image of the woman I have selected to use as the main subject of the image. I hope that the upscaling process doesn't deviate too far from the image I've constructed so far, as that means that the compositing in Photoshop will be far easier (and smoother) to get my desired result.

The first iteration of the upscale looks nothing like the woman in the first iteration of the image, which is disappointing.

image.png

I tried again, and got a more pleasing result, closer to the original rendering.

image.png

I mirrored the image, and made two versions come together:

image.png

I followed that up by cleaning up the middle of the image with the content aware fill tool:

image.png

Here, there is much more work to be done in order to progress the work to completion; including the addition of the other, upscaled figure, differentiating the gaze between the left and right side of the frame in order to create disharmony.

I took a MidJourney candidate, cut her in half, then re-mirrored her to get the following result:

image.png

I'll want to narrow and clean up the face a little bit, but I quite like the artefacts of the fact that this is obviously mirrored. It adds to the context and meaning of the work, in my opinion. Anyway, enough rambling, time to complete the tedious stuff in Photoshop to complete my vision, properly. :D

image.png

This is the composition nearing completion, but there's still a bunch of stuff to work on, namely the shape of the face of the center figure, followed by some dodging and burning.

Pretty obvious differences are evident with the change in the central figure, witnessed below:

image.png

There are still some dodging and burning (and frequency separation) steps that I want to take before I'm willing to call this image complete, so I will be working on them next.

A few adjustment layers, some more colour modification and very minor, subtle tweaks, and I arrive at the final result:

Mirrors.jpg

Sort:  

Man, you really take generative art into another level!
Inspiring!

Hey, thank you! This is my intent! As a trained actual "artist" I love how this tool can take practice to an even higher level. I'm working on some other stuff at the moment, and I'm finding that the AI is teaching me about symbolism in ways I would have never represented specific concepts, and it is incredible.

I absolutely love it, with a deep, deep passion!

Please keep posting more of this kind of contents. I learned a lot from your detailed processes breakdowns.
I'm trying to sneak some time soon in my tight daily schedules to take a deep dive into Stable Diffusion. Do you have any advise for newbies like me?

Look up a desktop application called "Visions of Chaos" and follow the guide on the website where you download it from for starters.

Interesting!
I just look it up, unfortunately I'm on my Macbook Air. I will check back when I can utilize my gaming PC soon.
Thank you very much 👍

hoping to see what you can come up!

Beautiful