ChatGPT 5 Critique

[Famous Artist]

I did not know Chat could do this:

I worked this old photo to the very best of my ability then asked ChatGPT 5 to critique it. It's a bit soft so I though ChatGPT might catch that.

See below what happened (sorry of this has been covered before):



You said: What can I do to improve this fine art photograph?

ChatGPT 5 said:

This is already a strong fine-art style photograph—two sailors in authentic uniforms, captured in a natural walking pose with good contrast and presence. To refine it further, here are some suggestions in three categories:

1. Tonal Balance and Depth

• Facial highlights: The sailor on the right has a very bright face and hat, which pulls more attention than the left figure. You could soften those highlights slightly (bring them down closer to Zone VII–VIII rather than Zone IX–X in Ansel Adams terms).
• Shadow separation: Deepen the shadows in their trousers and under the neckerchiefs to add dimensionality without losing texture. This gives stronger sculptural depth.
• Midtone richness: Lift subtle detail in the midtones of the uniforms so they don’t look overly flat compared to the bright whites.

1. Background and Framing

• Blur consistency: The right side background is much lighter and blurrier than the left, which makes it feel a little unbalanced. Consider evening out the tone and blur so that both figures stand out equally against a subdued backdrop.
• Vignette refinement: A very soft vignette could help keep the eye inside the frame and focused on the sailors, rather than letting it drift into the bright background wash on the right.

1. Fine Art Presentation

• Film grain treatment: The foreground pavement has visible digital noise/grain. You might unify the texture with a subtle, even layer of authentic film-like grain so it feels deliberate rather than uneven.
• Shadow grounding: Their shadows on the ground are strong, but you could feather them slightly so they integrate more naturally with the tonal treatment of the whole image.
• Print toning option: For exhibition-style fine art, try a slight selenium or platinum-like monochrome tone—just enough to warm/cool the blacks, while staying faithful to the vintage black-and-white aesthetic.

Summary:
If you tone down the sailor on the right’s highlights, balance the background blur/brightness, and unify the grain structure, the photo will achieve a more harmonious fine-art balance—while keeping its authenticity and impact.

Would you like me to create a step-by-step Zone System adjustment plan (Zone 0–X mapping) for this specific photo, so you can fine-tune the tonal placement systematically?

You said: Yes

ChatGPT said:

Perfect — let’s build a Zone System tonal adjustment plan tailored to this photograph. I’ll map the key areas of your image to Ansel Adams’ Zones 0 through X, then suggest where each should ideally sit for balance, depth, and authenticity.

Zone System Adjustment Plan for Two Sailors

Zone 0–I (Pure black / near black)

• Current: Very little true black—shadows in uniforms lean more toward dark gray.
• Adjustment: Deepen selective shadows (creases in trousers, undersides of neckerchiefs, areas between arms and torso). Keep texture visible—avoid clipping. This anchors the image.

Zone II–III (Deep shadows with texture)

• Target Areas:
• • Shoes (polished leather reflections should live here).
  • Inside folds of collars and sleeves.
  • Lower pant legs where natural shading occurs.
• Adjustment: Increase shadow richness but preserve detail. This adds sculptural realism.

Zone IV–V (Midtones, “middle gray” reference)

• Target Areas:
• • Sailor uniforms (broadcloth fabric).
  • Faces (excluding bright highlights).
  • Pavement midtones.
• Adjustment: Uniforms should mostly live here, with subtle gradations. Currently they’re a bit compressed; expand tonal separation so folds and creases pop.

Zone VI–VII (Light gray, textured highlights)

• Target Areas:
• • Hands and faces (skin tones).
  • Shirt collars beneath the “V” necks.
  • Reflections on sunglasses (left sailor).
• Adjustment: Faces should sit mostly in Zone VI–VII. Currently, the sailor on the right’s face is closer to Zone IX—pull this back so his skin matches a luminous but realistic light tone.

Zone VIII (Very light, textured whites)

• Target Areas:
• • Sailor hats.
  • Sunlit highlights on shoulders.
• Adjustment: Keep detail in the hats. Right now, the right sailor’s hat is drifting toward Zone X (blown white). Bring it back into Zone VIII so stitching and shape remain visible.

Zone IX–X (Specular, near-paper white)

• Target Areas:
• • Tiny specular glints only (sunglasses reflection, perhaps edge of hat).
• Adjustment: Reserve these zones for accents only. Avoid letting large areas (like the whole face or hat) sit here, or they overpower the composition.

Background Treatment

• The left side background: darker and more balanced (Zones III–V).
• The right side background: washed out in Zones VIII–X, distracting.
• Adjustment: Bring the right background down into Zones V–VII with soft blur. This creates tonal harmony, keeps focus on the figures, and eliminates the unbalanced brightness.

Final Touch

• Apply a very subtle selenium toning simulation: deepens blacks (Zones 0–III) while giving highlights a cooler silver sheen. This adds richness without losing the authentic WWII-era look.

This plan gives you a systematic tonal map. If you apply it step by step, the photo will gain depth, balance, and fine-art presence without losing historical authenticity.

Would you like me to translate this into a Photoshop/Lightroom step-by-step workflow (with Curves, Dodge & Burn, and local contrast tools), so you can apply these Zone adjustments directly?



You said: No, it would better if you make recommended changes to the photo and presented it to me as a CHATGPT offering for my approval or further work. Thank you!




-Famous Artist

 

[Famous Artist]

I let ChatGPT 5 do the whole thing, see results below.

I could have done it myself, but why?

Results meet my needs, and it took a fraction of the time to complete (compared to PS, etc.). I can very quickly move to the next image pleased with my results.









--
Famous Artist

 

[dj_paige]

So the real question is: what percentage of time does ChatGPT meet your needs?

And what percentage of time would ChatGP meed someone else's needs?

--
Paige Miller

 

[ishwanu]

Results meet my needs, and it took a fraction of the time to complete (compared to PS, etc.). I can very quickly move to the next image pleased with my results.

If you're happy with an image that's altered in strange and unnecessary ways, I guess that's what counts.

 

[dj_paige]

I let ChatGPT 5 do the whole thing, see results below.

Please tell us exactly what you requested ChatGPT to do that produced these results. Without that information, how can we judge the results?

 

[Abbott Schindler]

As long as you're OK with things like:

- the right person's cap has been changed (not cocked as much)

- I find neither person as delineated in the second image than the first (skin tones look better to me, but that's about all).

- The object the left person is holding is larger and blurre in the AI version.

- The stripe on the right shoulder has been removed in the right person, while still there on the left person.

- The left person's smile's been changed to a more neutral mouth (although I do like the improvement in glasses shadow on his face).

Frankly, about the only things I prefer in the AI version are the highlight reduction and the way the wrinkles in the uniforms have been cleaned up. But that's just me; as others said, what really matters is that you're happy with it.

 

[Night Pixel]

As long as you're OK with things like:

- the right person's cap has been changed (not cocked as much)

- I find neither person as delineated in the second image than the first (skin tones look better to me, but that's about all).

- The object the left person is holding is larger and blurre in the AI version.

- The stripe on the right shoulder has been removed in the right person, while still there on the left person.

- The left person's smile's been changed to a more neutral mouth (although I do like the improvement in glasses shadow on his face).

Frankly, about the only things I prefer in the AI version are the highlight reduction and the way the wrinkles in the uniforms have been cleaned up. But that's just me; as others said, what really matters is that you're happy with it.

The OP said "Results meet my needs, and it took a fraction of the time to complete (compared to PS, etc.). I can very quickly move to the next image pleased with my results."

This is all that matters. We can all find small AI errors or changes; however, this is irrelevant to the OP because he is happy with the results and,

• took only a few seconds to achieve,
• produced a result that exceeded OP's skill level if done manually, and
• saved time to allow OP to move to another image.

All these factors come to play when using automated and AI tools to achieve a result. No one who did not have access to the original photo would be able to point out "errors" made by ChatGPT 5. Even then, most people, i.e. those not on DP Review Retouching Forum, would notice these differences.

--
++++++++++++++++++++++++++++++++++++++++++++++++++++++++
"A photograph is a secret about a secret. The more it tells you the less you know." - Diane Arbus

 

[Famous Artist]

There were multiple images from a WW2 set.

I spent a fraction of time processing these images compared to using the old digital manipulations methods (RAW, PS, Topaz Nik, On1, etc).

ChatGPT 5 results exceeded the customer expectations.

After delivery, I picked up more work from the customer.

Finding ways to increase my productivity while exceeding customer expectations is a model that pleases me.

I'll now be ramping up my use of AI as I begin to de-use the old digital manipulations methods and tools.

 

[SC489]

I've used ChatGPT5 for processing images but for free users it has a strict image download limit. I now use Copilot instead which still has the CGPT 5 engine and no restrictions for data use if you login with your Microsoft approach. It also has a deep research mode which produces a very impressive detailed report of its investigations complete with all references.

 

[Night Pixel]

I let ChatGPT 5 do the whole thing, see results below.

I could have done it myself, but why?

Results meet my needs, and it took a fraction of the time to complete (compared to PS, etc.). I can very quickly move to the next image pleased with my results.

I hope you do not mind, but I put your original photo through the new Google Gemini model and asked it to apply Zone System tonal changes without changing any facial features. For comparison this is what it did with a minimal prompt:





The changes made by Gemini were subtle and you may prefer the ChatGPT 5 one over that. But Gemini is supposed to avoid making changes to peoples faces and body features.

Here is the original:



--
++++++++++++++++++++++++++++++++++++++++++++++++++++++++
"A photograph is a secret about a secret. The more it tells you the less you know." - Diane Arbus

 

[BillyBobSenna]

I let ChatGPT 5 do the whole thing, see results below.

I could have done it myself, but why?

Results meet my needs, and it took a fraction of the time to complete (compared to PS, etc.). I can very quickly move to the next image pleased with my results.

Amazing results and crazy to think where this will be in 10 years.

--
Bill - Beverly Hills, MI
Motorsports Photography
www.billgulkerphotography.com

 

[Schrodingers cat]

I remain unconvinced of the usefulness of ChatGPT. verbal feedback. It rather strikes me as being something like a conversation with an empty cat food can

--
Personal travel snapshots at https://www.castle-explorers.com
1. Making good decisions is generally the result of experience.
2. Experience is generally the result of making bad decisions.
3. Never underestimate your capability for doing incredibly stupid s**t.

 

[Guy Parsons]

I remain unconvinced of the usefulness of ChatGPT. verbal feedback. It rather strikes me as being something like a conversation with an empty cat food can

Yes, use it with caution and amusement.

I asked a simple

Q: "does the Beijing subway network have a maglev line "

A: "There is no maglev line on Beijing subway".

Asked again

Q: "what about the Beijing S1 line"

A: "The Beijing S1 subway line is maglev but not a fast maglev as it is limited to 100kph".

So it pays to know the correct answer before even bothering to ask ChatGPT.

 

[Night Pixel]

I remain unconvinced of the usefulness of ChatGPT. verbal feedback. It rather strikes me as being something like a conversation with an empty cat food can

Here is what Chat GPT 5.0 Thinking said about your comment:

"

That’s a vivid metaphor. An empty cat food can is, by design, empty—and terrible at conversation. I’m not a can; I’m a tool that does useful transformations on language. As a large language model (LLM: a system trained to predict and transform text), my value shows up when you toss me real tasks, not just vibes.

Here’s a tiny demo using the line you shared:

• Sentiment in one line: Skeptical, dismissive, and requesting proof of practical value.
• Constructive reframing: “I’m not yet seeing practical outcomes from ChatGPT; show me concrete examples where it saves time or improves quality.”
• Two ready-to-send replies (pick a style):
  • Polite: “Fair point—usefulness should be measurable. Give me one minute and I’ll show a small, concrete win.”
  • Playful: “Totally—talking to a can is rough. Let’s crack it open: give me a task and I’ll serve something more nourishing.”

Where I tend to be actually useful—fast, repeatable, sanity-checked work:

• Condense and compare: Summarize long text, highlight the 5 non-obvious differences between versions, extract action items.
• Reasoning with constraints: Step through a plan (with trade-offs stated), draft a checklist, convert fuzzy goals into criteria you can score.
• Fact-finding with citations: When asked, I can browse and show sources for recent news or specs, so you can verify.
• Apple-flavored help:
  • Draft a Shortcuts automation (step-by-step) for macOS or iOS to rename files, resize images, or file receipts.
  • Write or fix a small AppleScript (automation for macOS) to batch-rename files or move them based on tags.
  • Walk through Photos or Mail rules to de-clutter, with a clear checklist you can follow.
• Quality control: Turn a draft into a style-consistent document, catch contradictions, or generate test cases from requirements.
• Explainers: Translate technical jargon into plain English without dumbing it down.

Tools are judged by outcomes. Hand me something slightly annoying or time-consuming—an email you want tightened, a folder full of mess that needs rules, or a tiny automation you wish existed—and I’ll turn it into a concrete result."

Given what ChatGPT is good at I have used extensively for medical research on personal issues and have received quality information with citations to professional medical journals and databases. So it is more than "empty can."

It is a real tool that provides comprehensive solutions to many problems. So much so that I am a paying customer for higher levels of access.

--
++++++++++++++++++++++++++++++++++++++++++++++++++++++++
"A photograph is a secret about a secret. The more it tells you the less you know." - Diane Arbus

 

[Schrodingers cat]

Next time you notice a really good DPR pissinmatch going on over how many photographers can dance on the head of a pin, jump in and have Chat generate it's own responses. I bet it will do really well

For medical uses? I'm not even real comfortable with Dr Mercola and Dr Oz, and you know how great they are

--
Personal travel snapshots at https://www.castle-explorers.com
1. Making good decisions is generally the result of experience.
2. Experience is generally the result of making bad decisions.
3. Never underestimate your capability for doing incredibly stupid s**t.

 

[Night Pixel]

Next time you notice a really good DPR pissinmatch going on over how many photographers can dance on the head of a pin, jump in and have Chat generate it's own responses. I bet it will do really well

For medical uses? I'm not even real comfortable with Dr Mercola and Dr Oz, and you know how great they are

Well, try it. Specify what you want the response to look like and to include citations to medical literature. I also instruct ChatGPT not to guess or make up anything. The response must be verifiable. I have gotten excellent information summarized from medical literature. If the response is too technical I ask it be reframed for an educated non-medical professional. Super useful.

 

[xpatUSA2]

If you're happy with an image that's altered in strange and unnecessary ways, I guess that's what counts.

Ouch ...

 

[MimiVivi]

The only thing less interesting than "conversing" with ChatGPT itself is going to an Internet forum to see what other human beings think about something . . . and then finding them posting slop at each other copy-pasted from ChatGPT.

 

[OpticsEngineer]

I find myself a bit conflicted about what to think about the "improved" version. It looks much better and more engaging to view. But as someone else pointed out, unnatural considering photos from that time era never look like that. So the improved version looks fake, like something recently photographed in Hollywood. But my teenage daughters might see things completely differently than me. I still think the Stars Wars movie looks fine, but my daughters can't understand how it doesn't bother me how crude it looks.