As with architecture and interior design, fashion design represents fertile ground for exploration with AI and synthetic photography. Multiple instances of themed designs can be developed and modified, with variations in command specifications and photographic parameters. Here are some images I created modeling professional business attire. I appreciate how playing with this relatively simple theme leads to many images that showcase interesting ideas.
Midjourney makes it possible to rerun or iterate an image, exploring subtle or larger variations. This makes it well suited to generating and exploring ideas for further development with other means. Architecture and Design are fields for which digital image creation with A.I. is especially well suited. A great many of the world’s leading architects and designers include Midjourney or other A.I. program in among their tools.
The prompt for both of these images is /Imagine a Symmetrical picture, living room design, modern style, Minimalist style furniture, low saturation light gray, off – white gradient lighting, glass cabinet doors, LED lighting, Decorations, plates, jars, books, decorations, flower pots, ceramic art,greenery, high Resolution, 8K, Sony FE 85mm f/ 1 –
Here are two variations of a modern living room, Italian or European style. They represent a wealth of ideas about design and decor to prompt fourth development. (Click image to view full size.)
There’s a Midjourney command for generating mood boards. Here are two. The simple prompt here is /Imagine Moodboard 4×4 for design style ideas for modern family living area
I’ve been working with what has been called “Synthetic Photography.” As an old computer science and mathematics major, I have been intrigued by the development of generative A.I. The field of A.I. has been advancing by leaps and bounds, so that it is now possible to create photorealistic images. Synthetic Photography isn’t photography, but it isn’t just computer image creation. It’s a synthesis of photography and generative A.I. in a sense that I’ll discuss more fully in Blog entries. In the meantime, here are some examples.
/Imagine Advertising – style medium – close centered view photo of woman – model long hair in long gown on an abstract background shot with studio light 90mm f/ 4 – seed 1 – ar 16. 9 – v5
/Imagine Macro portrait a graceful blondewave 25 yearold supermodel with blue eyes smiling posing flirty on the beach and plush lips in the style of golden hour. Close up shot. Shallow field of depth –v 5.0 – Pan Down
/Imagine advertising-style medium-close centered view photo of smiling woman-model long black hair in dark red gown wearing gold necklace on an abstract background shot with studio light 90mm f/4 – seed 1 – ar 16.9 – v5
/Imagine Lifestyle, golden hour, full – length photo of young woman, short blonde hair in bright colorful dress with flowers design, wildflowers background, shot with natural light 210 mm f/ 22, right fill-light – seed 1 – ar 4:3 – v5 –v 5.0
(There’s a new image below following the text!)
Generative image AI works by analyzing millions of images and organizing them in a network. It assigns probabilities to different elements and relationships. When given a prompt, the AI uses this network to create an image related to the prompt. However, AI-generated content may not always be accurate or unbiased.
The irony in almost all A.I. systems is that while we generally know how it works, we don’t really understand why it works. But I think it’s fair to say that the A.I. searches its network for elements and connections that address the instructions provided in a prompt. Is this creative? Is it an example of real imagination? Perhaps not in the strictest human sense, but consider.
When it comes to creating art, like painting, there are different ways to be creative. If you’re new to painting, it’s common to start by copying others. But as you gain experience, you learn about colors, shapes, and textures. You also get inspiration from your own memories and the work of other artists. We store all of this information in your brain, forming a network of knowledge that helps you create.
When I paint, I gather information from my brain and use it to create my artwork. Similarly, AI uses its network of connections to generate content. This suggests that AI and our brains may share some understanding of creativity.
Comments welcome? What do you think?
Can’t have a blog post without an image. Here is a recent A.I. generated image.
I should be more or less routinely supplying the prompt for A.I. generated images. The prompt for this country scene in autumn is “a photorealistic image of a country scene in autumn, with a lake, and a family having a picnic alongside the lake, natural midafternoon light,lens 50mm f16 –ar 3:2 –s 200“
I have written that I prefer to use the A.I. to develop photorealistic images of things that might exist. I tend not to use the A.I. to create wild abstractions, fantasy images, or the like. Still, I’ve been moved to use the A.I. to imagine what certain things/concepts/ideas might look like. And I’m also intrigued to find out what the A.I. “knows” – or thinks it knows, and how it might imagine certain ideas. Here are a few examples of what I mean.
I asked Midjourney to produce an image that might imagine what the Roman Coliseum might have looked like in the days of ancient Rome. After much trial and error, numerous iterations, variations, and changes in the Imagine prompt, I ended up with this. It’s clear that Midjourney knows what the Coliseum is/was in Rome and has a general idea of what it looked like. The surroundings are pure A.I. invention, but it does capture the notion of ruins and/or construction. Many versions of this effort produced numerous “hallucinations,” which is when the A.I. produces something that just can’t be (like hands with 6 fingers, or walls that don’t touch the ground, or obscure things floating on the air.) It takes lots of trial and error to get out of the A.I. something that looks reasonable. Note that Midjourney does not know how many decks were/are in the Roman Coliseum.
Changing gears slightly, I was interested to what degree the A.I. could help interpret or realize an idea from literature. These two images were created in response to a prompt asking the A.I. to imagine something in the sense of the opening lines of T.S.Elliott’s Love Song of J.Alfred Prufrock: “Let us go then, you and I, when the evening is spread out against the sky like a patient etherized on a table.”
The first image is an early image produced by the A.I. It’s purely imaginative but at least it could bear some relationship to the text (without the A.I. having the slightest idea of what the text means). The second is a late stage image after numerous iterations and variations when the A.I. is allowed to introduce new content or interpretation.
This is a late stage image produced by the A.I. in response to the prompt imagining the well-known line of John Donne’s, “No man is an island, entire unto himself.”
In the next post, I’ll discuss some ideas about creativity and imagination.
I love the rich textures, the colors, the depth achieved by Midjourney in the field of wheat here.
I want to recommend Phil Steele, whose online course about Midjourney is the best, fastest, most easily navigated introduction to Midjourney. It’s a good introduction, and it then gets quickly into more advanced features to provide a solid beginning with Midjourney and Synthetic Photography.
There are at least two ways that A.I. imaging and traditional photography interact. The first I have already discussed. It’s useful, and sometimes necessary, to bring a traditional photographic sensitivity to creating images. Knowing about composition, camera and lens mechanics, exposure and lighting helps with giving Midjourney instructions to produce something like you envision.
The second is that the images digitally created by the A.I. need not be seen as final and definitive. The images can still be edited, in big or little ways, with any photographic editing programs like Lightroom. (I am devoted to Capture One.) Textures can be added or altered, lighting and exposure can be modified. The overall “look” of the image can be significantly tweaked – or changed altogether.
Here is another recent image in the florals category, irises at sunrise.
I have long admired the photography of Robert Mapplethorpe, especially his flowers. I have tried to replicate this work, capturing the control, creativity, and sensuality implicit in his works. Mostly unsuccessfully. Now, with the aid of generative A.I. I can come up with something close.
It’s not a photograph, and I would never claim it to be. It’s a digital image creation. It’s developed with an image A.I. program. (I use Midjourney, one of the most popular new image A.I.’s, and I think one of best both in terms of the range of what might be possible and the relative ease of use.) Note that I don’t try to create fantasy images or extreme abstract images. Rather, my aim is to combine a sensitivity to the conventions and mechanisms of traditional photography with the generating ability based on millions of image “learned” by the A.I. to create photorealistic images.
For example, here are the instructions (a “prompt”) passed to Midjourney to create the calla lily above: Closeup photo of white calla lily lying on table,sharp, black background,studio lighting from front, film Kodak T-max 400, lens 105mm, f16. The command that tells Midjourney to process this prompt is /Imagine. I think that command is apt and conveys a lot of meaning if we think about it. In essence, both of us, I and the A.I., are engaging in a kind of imagination. I’m trying to imagine the descriptions, along with parameters and other inputs, that may get me the image I want. The A.I. is being asked to imagine the kinds of images that would correspond to the prompt.
Floral and botanical images have long been a mainstay of my photographic interests. Here are some examples of floral images I have generated with Midjourney.
More examples of some of my Midjourney adventuring are in further blog posts along with a discussion of creativity and more about A.I. image processing. (Note: If you’ve come here direct from the blog [If you subscribed to the blog, this is the page you’ll land on] don’t miss the images on the site front page, here.)