What I would like to see out of Photo AI.

clee01l

Senior Member
Lightroom Guru
Premium Classic Member
Premium Cloud Member
Joined
Jun 20, 2009
Messages
22,107
Location
Houston, TX USA
Lightroom Experience
Power User
Lightroom Version
Cloud Service
I take lots of photos of the same subject. Often I will apply Denoise to the result. I end up with perhaps a half dozen or more of virtually the same image. All of them are technically acceptable. Some might be slightly focused on a different part of the image.

I would like an AI app or LrC function to evaluate these photos and pick the one that ticks the most check boxes. Surely this does not appear to be an unreasonable thing to ask of this AI tool?
 
I take lots of photos of the same subject. Often I will apply Denoise to the result. I end up with perhaps a half dozen or more of virtually the same image. All of them are technically acceptable. Some might be slightly focused on a different part of the image.

I would like an AI app or LrC function to evaluate these photos and pick the one that ticks the most check boxes. Surely this does not appear to be an unreasonable thing to ask of this AI tool?
How about a version of that AI tool that shows the differences between 2 or more (how many more?) photos. Not just the image coverage differences, but also exposure, and other items? I have the Compare mode in mind here.
 
How about a version of that AI tool that shows the differences between 2 or more (how many more?) photos. Not just the image coverage differences, but also exposure, and other items? I have the Compare mode in mind here.

An Adobe AI Compare mode might be just what I’d like to see. Especially if it would point out the exact pixels in focus. I’m not concerned about exposure as those settings are often identical across the images that i am trying to compare. Besides I can “fix” exposure issues in Lr when shooting RAW.


Sent from my iPad using Tapatalk
 
I don't think it's unreasonable at all, and should be done on those big computers in the cloud! In fact, for a few years there has been a Best Photos feature in LrWeb, a proof of concept that has gone nowhere, but I came home with 500+ photos on Saturday and would have loved to leave LR to find all the ones that are out of focus, apply compositional logic like Best Photos to point out the likely favourites etc.

I'd also like AI to learn my editing style and feed it into the Auto button or a preset. Can't it understand that (say) for images that are 1 stop underexposed I typically increase Exposure .25 and Shadows .5, and then do those steps for me? And can't it see that with certain types of images, I usually do certain things? So let's say it sees that with high key or low key images (it's got to recognise those) I often add a little positive or negative vignetting. Or where there's a lot of sky, I typically add a sky mask with a subtracted linear grad, and drop the Highlights and a little Exposure. Do what it thinks I would do with these images.

But from a very different angle, I'd like to see more disclosure of AI changes to images. Here I'm thinking more in terms of Photoshop and the Generative Fill, but I'd like to see an "AI fakeness" measure that reports how much of the image has been AI-generated. In simple terms, this might say that 1% or 35% of the pixels are faked, for example, taking us beyond the crude boolean. That 1% might be a relatively-innocent edit and the image may be judged to remain a "photograph", while the latter might tip it into being a "promptograph" and unusable in many environments while no doubt welcomed in other circles. A camera club might make a rule that 0.5% is acceptable for one competition and 5% for another, while a news agency might use 0.1% to trigger human review to verify that the photographer has only edited dust spots or phone wires, for example, and hasn't added or removed something that might change the photo's meaning. I don't know if you've yet experimented with Content Credentials in Photoshop but if you load CCA-tagged files into the online Content Credentials tool you can identify the presence of AI-generated content, but not the degree. What one does with that knowledge is another matter, of course.
 
Since most of my Photos are for personal consumption and I have the integrity to not submit and AI "improved" photo for competition, Most of my AI interest is toward comparing similar photos to use AI criteria to help me choose the "best" of the Lot.

On another note, I really have an appreciation for PS Generative expand for improving an image composition for fit a different aspect ratio. These derivatives I append a "GF" to the file name so that I can remember.
 
An Adobe AI Compare mode might be just what I’d like to see. Especially if it would point out the exact pixels in focus. I’m not concerned about exposure as those settings are often identical across the images that i am trying to compare. Besides I can “fix” exposure issues in Lr when shooting RAW.


Sent from my iPad using Tapatalk
I follow the Joel Satori philosophy of "work the scene" which often involves shooting different apertures for different DOF. Any AI for me has to be able to distinguish out of focus in the subject plane from variations of DOF. Grading or judging DOF could prove interesting as well, and dependent on what is in the actual background. I suppose it could judge me on intent as well, knowing the f/22 should be different than f/2.8 based on my intent.
 
....

On another note, I really have an appreciation for PS Generative expand for improving an image composition for fit a different aspect ratio. These derivatives I append a "GF" to the file name so that I can remember.
Great idea. I/we are still trying to decide how to deal with this in one of our club challenges. It is usually a bit different than adding another object that wasn't originally present.
 
Great idea. I/we are still trying to decide how to deal with this in one of our club challenges. It is usually a bit different than adding another object that wasn't originally present.

My camera club has an unlimited category where any thing goes. For a long time other categories have a rigid rule about removing anything that was not in the original scene. This includes telephone wires and manmade objects in a nature photo. Only dust removal and minor cosmetic adjustments are permitted.


Sent from my iPad using Tapatalk
 
Since most of my Photos are for personal consumption and I have the integrity to not submit and AI "improved" photo for competition, Most of my AI interest is toward comparing similar photos to use AI criteria to help me choose the "best" of the Lot.
I don't think it's a matter of integrity but of awareness. When you're in the middle of editing, it's easy to use a healing tool to fix a little dust spot and forget or just not know that it's generative AI. Yet that normally-permissible edit is going to trigger these boolean AI warnings. LR could embed that percentage in the metadata, for example.
On another note, I really have an appreciation for PS Generative expand for improving an image composition for fit a different aspect ratio. These derivatives I append a "GF" to the file name so that I can remember.
Not as far as I know. I think that's something you must have done when you used the feature.
 
Before focusing on all this AI goodness, I want to see Adobe address image management shortfalls. :D

I have to admit, some of the ideas here are much better than the recent updates to Lr or LrC.

Tim
 
It was an update on ChatGPT this night and I was told it creates photos much better
 
It was very interesting to read from Victoria about Firefly credits and how they can be used and in the future paid for. Did not even know they existed, but I am sure what ever plan they come up with will tailor my use in the future!
 
All these companies keep making AI tools that do the easy stuff.

I want an AI that wakes up at 2AM, drives my Jeep two hours to the trailhead, then hikes another two hours in the dark to the remote location I tell it, where it then uses my gear to capture an award-winning sunrise. Then it returns and wakes me up around noon so I can download the images and do the processing myself using non-AI methods.

C'mon Adobe... quit messin' around.
 
Back
Top