NikonGear'23
Gear Talk => Camera Talk => Topic started by: Michael Erlewine on January 30, 2025, 18:47:17
-
Since I write a lot of essays and articles and post to some 20,000 or so readers, I like to include an image along with the text to set the mood or reflect the topic. This can be a photo I take or a photo I license for use, and in the last couple years this has turned into a graphic image created by AI through (in my case) AI graphic software called “Midjourney.”
For these articles or essays, most often I am not looking for personal memorials, like snapshots of family or events, but rather by illustrations that evoke moods or feelings.
Although I always state if an image is AI prompted, most or many of my readers continue to see this image as a photograph, per se and congratulate me for the ‘photo’ even though I always post this disclaimer.
[Midjourney graphic prompted by me.]
I see in the future of photography two main roads or paths, taking photographs for family and personal documentation or memories, like snapshots, and impersonal illustrations that eventually will be better made though AI photography, where the supercomputer becomes effectively the camera whenever the goal is illustration rather than the memorial of something. Of course, I will continue to take traditional photos as I always have, with camera and lens, and just for the ‘art’ of photography.
It's like the question, which is more important, the razor or the shave? The shave is what we are after, not what kind of razor is used to get a shave.
It’s the same with photography. Which is more important, the image or the camera (how we made the image). I know photographers are having a fit about AI images just now, but that will fade out, IMO.
It’s the image that is important and not ONLY how we make it. Having made about 15,000 AI graphic images to date, I well know that prompting an image is not a walk in the park, but requires skill, hard work, and many iterations to get a good image.
Of course, a photography photo with camera and lens, by definition, will always be just that. However, after about 18 years of writing a daily article or blog, I tired of licensing photos, not because they were so expensive (although they can be), but because the selection was so difficult, and the time spent to find one so arduous.
When I first ventured into AI graphics, some years ago, the results were almost laughable, like cartoons, but as time progressed, they have gotten better and better.
And many early adopters that I knew were thrilled and considered their AI images works of art, and themselves artists, although I never saw that in their work. And, as time passed, most of these artists dropped out and disappeared. Instead, I always saw AI images not as art but as ‘illustrations’ and heaven knows I needed illustrations.
And so, fast forward to today, when AI is doing a pretty good job at illustration, I have to struggle to find one of my photos that is good enough AND will fit enough with the topic to be used. Sometimes I just post a photo I took anyway. LOL.
The takeaway for me is that AI is no threat to any kind of photography I do, because it does not contain any personal elements other than what I choose to put into it.
However, and this is important, more and more of the time AI images comes increasingly closer to satisfying what I consider important to have in a photo, however it is made.
And so, while I don’t fancy myself as an artist, I am becoming a better and better illustrator and AI images are doing things for me that I could never do (or would bother to do) myself. I wonder what readers here have to say about this.
Here are a couple of AI images used for an article on grains.
-
I also don't think AI images are a danger to photography. Just another tool to be used. Taking as an example the ones you show here, it is clear that you need not only skill in communication with the AI but also knowledge about your subject, about illustrations, graphic communication, elements of visual impact...and taste.
Please correct me if I'm wrong but the AI uses existing images to create a database /learn/ so basically you are remixing other peoples creation. I mena, AI will not know how to work depth of field if it didn't exist already, same with color palette, textures, compositional structures and so on.
-
Well put Paco - but how many of the people using AI to generate images will as you say understand the subtleties of the issues of colour palette, textures, compositional structures etc
Not so many I feel, so AI, as with smart phones and digital cameras, will enable more marginal images to be generated faster by the masses. Pity.
Maybe a new job opportunity for accomplished and experienced photographers with communications skills to become AI image specifiers?
I also don't think AI images are a danger to photography. Just another tool to be used. Taking as an example the ones you show here, it is clear that you need not only skill in communication with the AI but also knowledge about your subject, about illustrations, graphic communication, elements of visual impact...and taste.
Please correct me if I'm wrong but the AI uses existing images to create a database /learn/ so basically you are remixing other peoples creation. I mena, AI will not know how to work depth of field if it didn't exist already, same with color palette, textures, compositional structures and so on.
-
Yes, one has to learn HOW to prompt the bot, so to speak. One still has to have taste, skill, and all that it takes to use a camera and lens. Some godawful AI graphics are most common. There is no free lunch ever, anywhere.
It's another skil like learning Adobe products, most of which I already know how to use. I use Midjourney just as I do Photoshop, Illustrator, Audition, InDesign, Premiere Pro, After Effects, Acrobot Pro, Lightroom, Dreamweaver, etc., all of which I have learned how to do.
So decent looking illustrations as those I posted take skill and a good eye, in order to select the illustration appropriate for the texts or add is goes with.
If your are not skilled in Illustrator, Midjourney can do pretty much the same thing by prompting and selecting.
Here are a couple AI illustrations I created for various articles.
-
Impressive, Michael. If you fancy doing a live session one day, count me in!
-
Hello,
only my personal first impression:
I think photographic pictures allways should show something new, the ai pictures show something already seen, a little bit "artificial" and sterile, but also superreal and superperfect. At the moment they look like ai pictures based on "real" photos.
The question is: is ai capable to create something new?
-
Hello,
only my personal first impression:
in think photographic pictures allways should show something new, the ai pictures show something already seen, a little bit "artificial" and sterile, but also superreal and superperfect. At the moment they look like ai pictures based on "real" photos.
The question is: is ai capable of create something new?
The same i had been thinking! I think not since its using data which by definition is yesterdays'. Latest versions of editing software have been using AI as well, and if cameras have not been updated with AI techniques they soon will, I reckon!
-
Thus we no longer need to put any effort into making a photo, as the camera itself will know "best" .... oh Brave New World .... at last.
-
Yes, one has to learn HOW to prompt the bot, so to speak. One still has to have taste, skill, and all that it takes to use a camera and lens. Some godawful AI graphics are most common. There is no free lunch ever, anywhere.
Here are a couple AI illustrations I created for various articles.
I doubt in a serious Photo, these hands and this head would belong to the same person (male hands?) : Beside that skin of face and hands wouldn´t show an age difference of 30 years.
-
Thus we no longer need to put any effort into making a photo, as the camera itself will know "best"
there is no danger, no camera in the world can be creative, with or without ai, and ai software only can ruminate, making a photo only needs fun, enjoyment, easy playing, no hard work, playing not working!
-
Certainly, at least for me, AI graphics with Midjourney take far longer to create an image than with camera and lens. You have to really work at it. As mentioned, I have done about 15,000 of them and am still learning how to properly prompt an image. It's coming at image making from a whole new angle.
You can even choose a camera like a Nikon D850 and a particulat lens to have your image model.
-
Certainly, at least for me, AI graphics with Midjourney take far longer to create an image than with camera and lens. You have to really work at it. As mentioned, I have done about 15,000 of them and am still learning how to properly prompt an image. It's coming at image making from a whole new angle.
It looks hyper-perfect and surreal, like a rendering/composing, a little bit mask-like and liveless, but not unaesthetic
-
The upper and lower parts of her face are disconnected. That makes the appearance quite artificial.
-
The question is, what these pictures have to do with "Camera Talk", pictures without cameras
-
It's like the question, which is more important, the razor or the shave? The shave is what we are after, not what kind of razor is used to get a shave.
It’s the same with photography. Which is more important, the image or the camera (how we made the image).
[/quote]
I don't think it is the same with photography, at least not the sort of photography that NG members and millions of other enthusiasts do. I think we do photography to create images by using photographic techniques, not just to have an images.
When photography was invented, some said that it would replace painting. Spoiler alert, it didn't. That is because people wanted to produce images by painting, and people wanted to acquire painted images whether or not a photograph of something similar was available.
I can see that AI images can be useful for generic purposes, such as advertising.
Computers can defeat the best chess players, but people still pay to watch Magnus Carlsen play chess, and not just in the hope of witnessing a meltdown over jeans.
Horses and cars can go faster than people, but the Olympics still attract millions of viewers.
-
Can you call it a photograph if no camera has been involved at any time?
Then you could start in a drawing or paint program with a white sheet and define each pixel by hand (it will take some time).
There is a finite number of "photographs" possible.
I have attached one of them just to push it to the extremes.
Is this a photograph and if it is not why are the AI-generated photographs?
Is it just if you accept it as a photograph when you look at it?
My personal opinion in the moment is that a camera has to be involved but it can be difficult to prove if it is the case. I may change my opinion at some time.
Using a camera some very abstract photographs can be made. It is just light that has been captured in some way.
-
If the implication of 'using a camera' means an optical system records an image, it is entirely feasible to make a photograph without a camera. Or even AI :)
This is an image obtained by using neither camera nor lens ....
-
If it is not a camera then it is not capture of light?
Mathematical formulars, algorithms etc. can generate digital images.
Random numbers.....fractals....etc.
-
A link to fractals images:
https://stock.adobe.com/search?k=fractals&asset_id=191237819 (https://stock.adobe.com/search?k=fractals&asset_id=191237819)
-
If it is not a camera then it is not capture of light?
Mathematical formulars, algorithms etc. can generate digital images.
Random numbers.....fractals....etc.
The image is definitively a record of a light field, and in facts depicts a natural fractal phenomenon. It is NOT generated by any mathematical equation.
Neither a traditional camera nor a lens was used.
-
Ok, if it is a record of light then you can probably also define the "camera" at an abstract level?
Display of recorded light is a photograph?
-
'Photograph' as it is made by light striking a recording surface ....
-
AI will be much used for Illustration; especially illustrations for Advertising which attempt to pretend to be real photographs.
At the moment the tell-tale signs of the use of AI-generated images are seen in images of hands, feet, skin-texture, eyes and are clearly obvious in the unmatched lighting in different parts of the composite. This will improve with time.
My personal view is that I only want to use photographs which I have shot myself so I have never bought Stock photographs.
I may use AI techniques and editing to combine a better sky or background with a foreground object but all of the components of the final images will have been generated by my own cameras.