There are two different jobs hiding under photo editing prompts, and writing the same prompt for both is why people get poor results. One is asking for an edit recipe you then run yourself in Lightroom or Photoshop. The other is asking an image model to perform the edit on a file you upload.
They need different prompts. A recipe wants slider values, layer order and reasoning. A direct edit wants you to state what must stay unchanged before you say what changes, because the usual failure is the model rebuilding something you never asked it to touch.
The fifty below are split accordingly, with three further groups for the parts that surround the edit: working out what is actually wrong, keeping a set consistent, and preparing files for wherever they are going. For prompting technique across other kinds of work, see the main ChatGPT prompts guide, and ChatGPT covers what the model itself can and cannot do.
The Two Modes, and Why the Prompt Differs
Recipe mode: ask for values and order. Eight steps with a direction for each beats a paragraph of advice. Ask for the reasoning on each step and you learn something transferable to the next photograph.
Direct edit mode: protect before you change. Name what must stay identical first. Adding do not alter anything else sounds redundant and measurably reduces the amount of collateral redrawing you have to undo.
Diagnose before either. Most bad edits are fixes applied to the wrong problem. Ten minutes spent asking what is actually wrong saves an hour of adjusting.
Say where the image is going. Crop, export settings and sharpening all depend on the destination, and none of them can be chosen sensibly without it.
A request looks like this:
Make this photo look better.
The same ask as a prompt:
Look at this photograph and tell me the three things most responsible for it not looking right, in order. Name each as a specific technical issue rather than a feeling, and say which one to fix first. Do not suggest any adjustments yet.
The second one tells you what to do. The first one gets you a list of things to try.
Diagnose What Is Wrong
Before you adjust anything. These ten produce a description of the problem rather than a solution, which is the step almost everybody skips.
1. Why does this look off
Look at this photograph and tell me the three things most responsible for it not looking right, in order. Name each as a specific technical issue rather than a feeling, and say which to fix first.
2. Exposure read
Analyse the exposure in this image. Tell me where it is clipping in the highlights and where the shadows have gone flat, and whether the problem originates in capture or in processing.
3. Colour cast identification
Identify any colour cast in this image, name it by hue and say where it is worst, then tell me whether it is a white balance issue, mixed lighting, or a reflection off something in the scene.
4. Where the eye goes
Tell me, in order, the first three places a viewer's eye lands in this image. Then tell me whether that matches what the photograph is supposed to be about.
5. Over-edited check
Tell me whether this image looks processed, and if so name the specific giveaways: halos, crushed blacks, oversaturated skin, unnatural sharpening. Be blunt about it.
6. Composition audit
Assess the composition: what is in frame that should not be, what is cut off that should not be, and whether a different crop would improve it. Do not suggest filters or adjustments.
7. Skin tone accuracy
Assess the skin tones in this portrait. Tell me whether they read as natural, in which direction they are shifted if not, and whether the problem is global or confined to the shadows.
8. Sharpness diagnosis
This image looks soft. Tell me whether that is missed focus, motion blur, camera shake or diffraction, and explain how you can tell them apart from what is visible.
9. Compare against a reference
Here is my edit and here is a reference image whose look I want. List the specific differences in contrast, colour and tonal distribution, and rank them by how much each contributes to the gap.
10. What to leave alone
Look at this image and tell me which parts I should not touch. Most editing damage comes from fixing things that were already working.
Ask for the diagnosis with adjustments explicitly forbidden. Otherwise you get a fix before anybody has established what the problem was.
Edit Recipes to Run Yourself
Instructions you execute in Lightroom, Photoshop or Capture One. The model is not touching your file here, which means you keep full control and learn something you can reuse.
11. Lightroom recipe with reasoning
Give me a Lightroom edit for this image as an ordered list of adjustments, with a value or direction for each and one line on why. Start with the adjustment that makes the biggest difference and stop at eight steps.
12. Layer order in Photoshop
Describe a non-destructive Photoshop edit for this image as a layer stack from bottom to top, saying what each layer does and which are masked. Use no destructive operations anywhere.
13. Break a look into its parts
I want this image to look like [described style]. Break that look into its component adjustments - tone curve shape, saturation by channel, grain, sharpening - and give me the settings in the order I should apply them.
14. The curve, described precisely
Describe the tone curve I should apply, as points with input and output values, to achieve [effect]. Explain what each point does to which part of the image.
15. Colour grading in three zones
Give me colour-grading settings for shadows, midtones and highlights to achieve [mood]. Specify hue and saturation for each zone, and say which zone is doing most of the work.
16. Local adjustments plan
List the local adjustments this image needs as a set of masks: what each mask covers, what adjustment it carries, and roughly how strong. No global adjustments in this answer.
17. Retouching, restrained
Give me a retouching sequence for this portrait that removes temporary blemishes and leaves texture, pores and permanent features intact. State explicitly what should not be removed.
18. Recovering a difficult file
This image is underexposed by about two stops and shot at high ISO. Give me the recovery sequence in the correct order, and tell me at which step I should accept the noise rather than fight it.
19. Black and white by channel
Give me a black and white conversion for this image using channel mixing rather than desaturation. Specify the channel values and explain what each does to the subject.
20. Preset, then the exceptions
Turn the edit described below into a reusable preset, then list the three adjustments that should never be part of a preset and must be done per image.
Always ask for the reasoning alongside the values. Numbers alone are a one-off fix; the reasoning is what you carry to the next photograph.
Here the model does the editing. Every prompt states what must stay identical before it says what changes, which is the single habit that separates a clean edit from a redraw. Working image to image is where Nano Banana Pro and GPT Image 2.0 are worth comparing on the same file.
21. Background replacement
Keep the subject in the attached image exactly as is, including the edge detail in the hair. Replace the background with a seamless mid-grey studio backdrop and adjust the subject's edge lighting to match a single softbox from the upper left. Do not alter the pose, skin tone or clothing.
22. Object removal
Remove [object] from the attached image. Reconstruct what is behind it so the repair is undetectable, matching existing texture, grain, shadow direction and colour. Leave everything else exactly as it is.
23. Lighting change only
Keep every object, position and framing in the attached photograph identical. Change the lighting from overcast to hard late-afternoon sun from the left, with defined shadows and a warmer white balance. Add nothing and remove nothing.
24. Selective focus
Reduce the depth of field in the attached photograph. Keep [subject] sharp and blur everything else progressively by distance, with natural circular bokeh matching an 85mm lens at f/1.8. Preserve the original colour grade.
25. Extending the frame
Extend the attached image outward to a [ratio] frame. Generate plausible content continuing the existing scene, matching grain, colour grade and perspective exactly. Do not alter or rescale the original area.
26. Object insertion with matched light
Insert [object] into the attached image at [position]. Match the existing light direction, shadow softness and the reflection characteristics of similar surfaces already in frame. Change nothing else.
27. Colour correction, content locked
Correct the colour cast in the attached image without changing composition, sharpness or any content. Afterwards, tell me what you changed and by roughly how much.
28. Time or season change
Convert the attached scene from [current] to [target], keeping the composition, structures and framing unchanged. Change only what that shift would genuinely alter.
29. Product on a new surface
Place the attached product onto [surface] in a studio setting. Match perspective to a slightly elevated three-quarter camera, generate a correct contact shadow and soft reflection, and light with one large softbox from the upper left. Keep the product artwork and proportions untouched.
30. Style transfer, structure preserved
Re-render the attached photograph as [style], keeping the composition, perspective and every structural element in exactly the same position. Do not add or remove any objects.
Keep the original. Direct edits are not reversible in the way a Lightroom adjustment is, and the version you liked is often two generations back.
Consistency Across a Set
One good edit is easy. Twenty that look like they belong together is the actual job, and it is where most amateur work gives itself away.
31. Matching a batch to one reference
Here is my reference edit and five other images from the same shoot. For each, tell me the specific adjustments needed to match the reference, and flag any that cannot be matched because of how they were shot.
32. What changed between frames
These two images are from the same session and do not match. Identify what differs - white balance, exposure, angle of light - and tell me which one to correct towards.
33. A look that survives different lighting
I need one editing approach that works across images shot indoors and outdoors on the same day. Tell me which adjustments should be global and which must be done per image.
34. Sequence order
Here are twelve images for a sequence. Order them so the tonal and colour shifts feel deliberate rather than accidental, and tell me what rule you used.
35. Subject consistency across generations
Using the attached reference, generate three more images keeping [subject] identical in every respect. Change only the background and camera angle, and match the lighting exactly.
36. Building a house style
From these ten images I like, extract the consistent editing decisions and write them as a specification somebody else could follow. Include what is deliberately variable.
37. Catching the odd one out
Here is a set intended to look consistent. Identify the images that break the set and say precisely why, in technical terms rather than impressions.
38. A preset across different skin tones
This preset was built on one image. Tell me which of its adjustments will not translate well across different skin tones, and what should be changed per image instead.
39. Print versus screen
This set is going both to print and to a website. Tell me what will shift between the two, and which adjustments should differ between the two versions.
40. Documenting the edit
Describe the edit applied to this image precisely enough that somebody could reproduce it on a different file. Include what to change if the new file is exposed differently.
Edit the hardest image first. If your approach survives the worst file in the set, everything else falls into line behind it.
Preparing Images for Where They Go
The stage that gets skipped, and the one that decides whether anybody sees the edit properly. If you are generating rather than editing, our ChatGPT prompts for image generation and AI image prompts cover that side.
41. Crop for each platform
Here is one image. Tell me the best crop for a square feed post, a vertical story and a wide banner, and for each say what is lost and whether the image survives it.
42. Copy space by design
I need this image to carry a headline in the top left. Tell me whether it currently supports that, and if not, how to recompose or extend it so that it does.
43. Export settings
For [use case], recommend export settings: dimensions, colour space, compression and output sharpening. Explain which of these matters most for this specific image and why.
44. The thumbnail test
Describe how this image will read at thumbnail size and what will be lost. Tell me the one change that would most improve its legibility when small.
45. Alt text from the image
Write alt text for this image in under 125 characters, describing what a reader who cannot see it would need to know in this context. Do not begin with image of.
46. Accessibility check
Assess this image for accessibility: contrast of any text overlay, whether meaning depends on colour alone, and whether important detail is lost for a colourblind viewer.
47. Print preparation
I am printing this at [size] on [paper]. Tell me what to adjust for that paper's tonal range, and what will look different from the screen version.
48. File hygiene
List what should be stripped from this file before publishing and what should be kept: location data, camera data, copyright fields, colour profile. Give the reasoning for each.
49. Version naming
Propose a file naming and versioning convention for a shoot with multiple edits per image, several output sizes and two people working on it. Keep it short enough that people will actually use it.
50. The final check
Before I publish this image, list the five things I should look at one more time, ordered by how often each turns out to be the thing that was wrong.
Decide the destination before you start editing, not after. A crop and a sharpening amount chosen for nothing in particular will suit nothing in particular.
Using These Prompts in Chat Smith
The two modes suit different models, which is the practical reason for having several in one place. Recipe and diagnosis prompts are text work and go to whichever model you find most honest about problems. Direct edits go to an image model, and it is worth running the same file through two of them, because they fail differently and you will not know which handles your kind of image until you try.
One habit worth forming: after any direct edit, ask what did you change that I did not ask for. It catches the quiet alterations - a shifted crop, a smoothed texture, a colour that drifted - that are easy to miss on a screen and obvious in print. Chat Smith is free to try if you want to compare a few models on the same file first.
ChatGPT prompts for photo editing are text instructions that tell an AI image tool how to change an existing picture rather than create one from scratch. A workable prompt states what to keep, what to change, and the finish you want, such as a new background, warmer color grading, or a cleaner product shot.
The Chat Smith Editorial Team is a group of AI enthusiasts, researchers, and content creators passionate about making artificial intelligence more accessible and practical. Through the Chat Smith blog, we share the latest AI trends, tool reviews, industry insights, and actionable guides to help individuals and businesses get more value from AI. Our mission is simple: deliver clear, reliable, and easy-to-understand content that helps readers stay informed, productive, and ahead in the fast-moving world of AI.
Share this article
Related Articles
Level Up Your Work,One Click Away!
Everything you need to push projects forward is right at your fingertips.