Has anyone here used AI to help with composition or editing decisions and does it actually improve your work - what would you do

[Gareth5]

Not talking about generating images. Talking about using AI to analyse an existing photograph and give feedback on composition, light, colour grading direction, or what is working and what is not. I tried this with Claude last month on a set of landscape shots and got genuinely useful observations about where the eye was being led versus where I intended it to go.

Q: Is AI art feedback actually useful or is it just generic praise with caveats?

A: Depends entirely on how you prompt it. Generic prompt gets generic feedback. Specific prompt gets specific feedback. Asking it to identify the strongest compositional element and explain why the secondary elements either support or compete with it gets you something actionable

[SharpLantern]

I do this regularly with street photography edits. Ask it to compare two crops and explain the tradeoffs. About 60 percent of the time it surfaces something I had not consciously noticed

[TommyB_20]

The 60 percent figure sounds right to me. It is a useful second opinion not an authority. The same way you would value feedback from a technically literate friend

[Amy]

The risk is over-relying on it and homogenising your eye toward whatever the model has been trained to consider good composition

[Finley_19]

That is a real risk for developing photographers. If you are still forming your aesthetic you probably want human feedback from people with taste rather than a model trained on a broad average

[Paige_68]

For experienced photographers with an established vision it is less risky. You know when it is wrong and you can use it selectively

[Olivia78]

I asked Claude to describe what mood a series of portraits conveyed without telling it what I intended. The gap between what I thought I captured and what it described was instructive

[Chris27]

That gap revealing exercise is brilliant. I am going to try this tonight with a set I have been unsatisfied with but could not articulate why

[Lazy Sentinel]

Post the results back here. Would be interesting to see how it translates to an actual editing decision

[JayJ]

Will do. I am curious whether the AI description matches the feedback I got from a print club last week that I also could not fully process