ChatGPT Image Detail Prompts: 10 Before and After Examples

A soft-looking image can need several different changes: clearer boundaries, stronger midtone separation or more visible surface texture. A useful ChatGPT image detail prompt names that change and the details that must survive it.

This guide covers ten common shorthand labels, with an original sample image and an actual refined result for each one. Use the full instructions below; labels such as /sharpen and /texture are informal reminders, not documented ChatGPT commands that unlock dedicated controls.

How these examples were made: We generated ten fictional inputs, then supplied each actual image to Codex's built-in image generator with its matching edit instruction on 8 November 2026. The underlying model name was not exposed. These are recorded generated-image edits, not tests of the ChatGPT website, real-camera restoration or recovered factual detail. The input prompts requested moderate softness; several inputs already retained substantial texture. Differences are visual observations, not measured quality scores.

How to use the prompts

1. Attach one image you own or have permission to edit in a tool that supports image inputs.
2. Choose the closest task below. Copy its full edit prompt, then replace subject-specific details with the ones visible in your upload.
3. Make one focused request. State what should change and name the face, object geometry, texture, lighting or framing that must stay.
4. Compare the original and result at the same scale. Check object counts, outlines, small markings and any lettering before accepting it.

OpenAI's image-generation guidance recommends targeted revisions and explicit preservation instructions. Its image-prompting guide also advises inspecting the result against the brief. Those principles inform the workflow here; they do not guarantee an unchanged subject.

Keep the original file. For a further attempt, start from it when you want a clean comparison. If you continue from an edited output, repeat the preservation details and check whether unwanted changes accumulate.

One reusable template, ten focused examples

Open the editable image-detail refinement prompt to change the refinement and preservation fields. Its default is the owl sharpening example. All ten command labels and their sample pairs are also included on the prompt page.

Edit the supplied image. Focus: [refinement]. Preserve [preserve]. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

The following sections show the exact expanded edit requests used in this test. Each pair places the generated input first and the edited result second; enlarge either image to inspect it. The sample-generation prompt is included so you can build a fresh practice input, though a repeat run will produce a different image.

1. /sharpen: Sharpen soft edges

Choose this when an otherwise usable subject looks a little soft. Name the important boundaries and leave a deliberately blurred background alone.

Edit prompt used

Edit the supplied image. Focus: Improve moderate edge sharpness around the owl's eyes, beak and feather boundaries. Keep the feather pattern natural; avoid white halos, exaggerated eye contrast and invented feather markings.. Preserve the owl's eye colour, beak shape, head outline, feather markings, perch and background. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

The eyes, beak and feather boundaries look more distinct, and the owl and perch stay recognizable. Fine feather edges have been reinterpreted; this is a generated refinement rather than recovered wildlife detail.

2. /enhance-detail: Enhance broad scene detail

Use this for a broad scene whose larger surfaces need more definition. Give the model landmarks to preserve rather than asking for unlimited detail.

Edit prompt used

Edit the supplied image. Focus: Increase the visibility of broad rock, snow and tree detail with restrained local definition. Preserve the existing mountain outline and snow boundaries; do not add peaks, trails or trees.. Preserve the mountain silhouette, snow boundaries, tree positions, valley layout and sky colour. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

Rock and snow divisions are clearer, with much more definition in the foreground conifers. The broad mountain silhouette stays similar, but fine snow and tree patterns change. Use this for a creative landscape, not a factual record of terrain.

3. /clarity: Improve midtone clarity

Choose clarity when the image feels flat rather than simply out of focus. Ask for moderate separation in the midtones, then check bright walls and dark edges.

Edit prompt used

Edit the supplied image. Focus: Improve midtone separation and restrained local contrast in the stone facades and street lamp. Keep highlight and shadow detail readable. Avoid an HDR look, dark halos and dramatic colour changes.. Preserve the street perspective, lamp geometry, window positions, facade colours and lighting direction. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

The lamp and pale stonework have stronger separation while the street keeps its gentle light. The result also defines some paving and wall texture. Check distant doors and windows when you need architectural accuracy.

4. /texture: Bring out surface texture

Use a texture instruction when the surface is already visible. Describe the material and protect its existing pattern instead of asking the model to invent a richer one.

Edit prompt used

Edit the supplied image. Focus: Make the already visible wood grain and growth rings easier to distinguish. Preserve the ring arrangement and existing cracks. Avoid carving new grooves or adding a rough artificial overlay.. Preserve the log outline, ring arrangement, cracks, wood colour and lighting direction. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

The growth rings stand out more strongly and the main cracks remain recognizable. Some fine ring lines are redrawn and the wood looks warmer. Those changes show why added texture should not be used to count real growth rings.

5. /micro-detail: Refine small visible features

Use this for small features that are present in the input, such as droplets or leaf veins. Count repeatable objects before and after the edit.

Edit prompt used

Edit the supplied image. Focus: Refine the visible leaf veins and edges of the five existing water droplets with subtle definition. Do not add droplets, change their positions or invent a new vein network.. Preserve the leaf shape, central vein, five droplet positions and sizes, green colour and background. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

All five droplets remain visible in the same broad arrangement, and the leaf veins and droplet rims are clearer. The finer vein network and highlight shapes change, so the edit is not a microscope-style recovery of the original surface.

6. /edge-detail: Define a subject boundary

Choose edge detail when the subject blends into its background. Specify the boundary that needs attention and preserve softness elsewhere.

Edit prompt used

Edit the supplied image. Focus: Improve the visible boundary definition of the cat's fur, ears and whiskers against the background. Keep the fur soft and natural. Avoid double outlines, halos, extra whiskers and sharpening the background.. Preserve the cat's face proportions, green eyes, stripe pattern, ear shape, pose and background. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

The ear outline, fur boundary and whiskers are more distinct against the grey background. The tabby pattern and pose remain recognizable, but individual hairs and whisker paths differ. Judge the silhouette without assuming pixel-level preservation.

7. /face-detail: Refine a face gently

Use a gentle face instruction when identity matters. Preserve proportions and expression explicitly, and compare the face with the original at a useful viewing size.

Edit prompt used

Edit the supplied image. Focus: Gently improve the visibility of existing eyes, eyebrows, freckles and skin texture. Preserve the person's facial proportions, apparent age and expression. Do not beautify, smooth the skin, enlarge eyes or invent freckles.. Preserve the fictional person's identity, face proportions, apparent age, expression, freckles, hairstyle, clothing and background. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

The eyes, brows and skin texture read more clearly while the face, expression and hairstyle remain recognizable. Fine freckles, pores and hair strands shift. This fictional portrait demonstrates the risk of treating a sharper face as a verified likeness.

8. /fabric-detail: Clarify fabric weave

Name the weave, seam or folds you need to see. A fabric-specific instruction helps distinguish useful textile detail from a generic sharpening effect.

Edit prompt used

Edit the supplied image. Focus: Improve the visibility of the existing diagonal denim twill and seam stitching. Keep the fabric flexible and matte. Do not add stitches, change the weave direction or make the cloth look metallic.. Preserve the fold shapes, seam path, stitch arrangement, button geometry, indigo colour and lighting. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

The denim twill and stitching are easier to see, and the fold and button arrangement stay broadly consistent. Fine thread placement and button marks change. Inspect the seam closely before using a generated edit to represent an actual garment.

9. /material-detail: Keep different materials distinct

Use this when several materials share a frame. Describe how each should read, such as transparent glass and brushed metal, rather than sharpening every surface equally.

Edit prompt used

Edit the supplied image. Focus: Make the visible brushed steel, clear glass and stone surface easier to distinguish through restrained texture and reflection definition. Keep glass transparent and metal reflections plausible. Do not add scratches, logos or new components.. Preserve the carafe geometry, glass transparency, metal finish, reflection direction, counter and background. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

The steel filter, glass carafe and stone counter look more distinct. Transparency and the main reflections remain readable, but the filter surface and tiny glass details change. A material prompt can also reinterpret the finish of a real product.

10. /crisp-photo: Give a whole scene a restrained finish

Use this as a restrained whole-image finishing request. It overlaps with sharpening and detail enhancement; the label itself does not select a separate tool.

Edit prompt used

Edit the supplied image. Focus: Apply a restrained overall crisp finish: slightly clearer lighthouse edges, readable rock texture and balanced midtone definition. Keep the sky and water smooth. Avoid oversaturation, crunchy detail, halos and extra scenery.. Preserve the lighthouse shape and position, red cap, rock arrangement, horizon, sunset colours and composition. Keep the same subject, camera angle, crop, composition, colours and lighting except for the requested local definition change. Make a restrained, natural-looking refinement; do not redesign the scene or add objects. Treat uncertain fine detail conservatively rather than claiming to recover it. Return one image in the same landscape 3:2 composition, with no text, labels, border or watermark.

What changed

The lighthouse edges and foreground rocks are clearer while the sky stays comparatively smooth. The tower, red cap and sunset remain recognizable. Fine rock texture, grass and masonry change, so the result is a creative finishing example.

Choose the smallest change that solves the problem

For a soft outline, begin with the sharpening or edge-detail wording. For flat stonework, try clarity. For a visible weave or grain, name the surface. The labels overlap: /sharpen, /enhance-detail and /crisp-photo can lead to similar requests. Their practical value is helping you describe a target, rather than acting as separate filters.

In our ten pairs, the broad scenes stayed recognizable, but the fine texture often changed. The mountain's snow divisions, the leaf's smaller veins and the portrait's freckles show why a visually crisper result needs inspection. A pleasant-looking result can still misrepresent a product finish, a person or a physical surface.

If the task requires an exact historical record, scientific detail, garment construction or product specification, do not use these examples as proof that missing detail can be recovered. Keep factual work anchored to the original and use an editing process whose changes you can inspect and control.

Try the workflow with your own image

Start with the editable refinement template, name one visible problem and list the details that matter. Reduce the request if the model changes too much. Changed inputs, repeat runs and other tools may behave differently from these recorded examples.

Techotick used AI assistance for research, writing and editorial revision. All 20 example images are generated media created for this guide. The before/after comparisons show actual edits of synthetic inputs and do not establish real-photograph restoration performance. No sponsorship or affiliate links are included.