30 ChatGPT Photo Enhance Prompts: Blurry to Clear and HD Without Changing the Face (2026)

A ChatGPT photo enhance prompt is an instruction attached to an uploaded photo that tells ChatGPT Images 2.0 to make it sharper, cleaner, or brighter while a keep list holds the face in place. ChatGPT regenerates a plausible sharper version of your photo instead of recovering detail the camera never captured, which is why faces drift when the prompt leaves them open. The 30 prompts below close that gap.

What Makes ChatGPT Photo Enhance Prompts Work

An enhance prompt asks for the smallest change of any edit, and it is the one most likely to redraw a face. ChatGPT Images 2.0 has no hidden pixels to pull out of a blur. Without the selection brush it regenerates the whole frame from your photo plus the instruction, and "make it clear" is read as permission to decide what a clear version of you looks like.

Say what the model is allowed to invent. A blurred face holds no eyelashes. Ask for "sharp eyes" and the model draws the eyelashes it thinks belong there, and the result is a sharper photo of someone slightly different. Say "recover detail only where the original shows it" and name the parts that may be sharpened: hair edges, fabric, the background.

Name the features so the face is on the record. "Keep my face" is too loose to hold. "Keep my face shape, the spacing of my eyes, my nose width, the shape of my lips, the mole on my left cheek, my hairline, and my expression" gives the model a checklist, and it preserves what it is told to notice. Every prompt below carries that list; the Keep-the-Face Set extends it.

Ask for up to 2K, since that is the ceiling. Output tops out at 2K resolution. A prompt that asks for 4K gets a 2K image anyway, so the number buys nothing. Ask for "up to 2K, same aspect ratio" and you get a predictable file every time.

Brush around the face when the fix is local, and forbid the quiet edits with a negative prompt. The selection brush regenerates only the painted region. Paint the background, the clothing, or the blurred corner, leave the face unpainted, and the face cannot change because it was never redrawn. For global fixes such as exposure and white balance the brush is the wrong tool, so describe the change and rely on the keep list.

Know when a dedicated upscaler is the better tool. For a pure resolution bump on a photo that is otherwise fine, a dedicated upscaler keeps the pixels more faithfully, because it enlarges what is there instead of regenerating it. ChatGPT earns its place when the photo also needs light, color, or clean-up fixes, since it does all of them in one pass under one keep list.

This post is a spoke of 40 ChatGPT photo editing prompts, which covers backgrounds, restyles, restoration, and wardrobe. If none of the 30 below fit your photo, the AI prompt generator writes an enhance prompt with the keep list built in from a plain-English description of the problem.

Every prompt below assumes one workflow. How to enhance a photo in ChatGPT without changing the face, in six steps:

Unblur and Sharpen (1–6)

A ChatGPT photo enhance prompt for blur works best when it says what kind of blur it is, since motion blur, a focus miss, and compression leave different marks and get different fixes. Name the direction of the smear or the plane that is sharp, name what may be sharpened, and forbid invented detail. The face stays on the keep list in every one.

1. Fix Motion Blur

This photo has motion blur from camera shake, smeared [left to
right]. Reduce the smear and bring my hair edges, clothing, and the
background back to sharp, recovering detail only where the original
shows it. Keep my face shape, eyes, nose, mouth, skin tone,
expression, pose, and framing exactly as they are. Do not invent
eyelashes, pores, or texture. No halos, no plastic skin. Up to 2K,
same aspect ratio.

What you'll get: A steadier-looking frame with the face you started with.

Variation: If only the subject moved, add "the background is already sharp; leave it alone."

2. Rescue a Focus Miss

The camera focused on the [background / my shoulder] and my face is
slightly soft. Shift the apparent focus to my face: sharpen my eyes,
brows, and hair edges to match the sharpness of the [background],
and leave the rest of the frame as it is. Keep my face shape,
features, skin tone, expression, and pose exactly as they are. Do
not redraw my eyes or add detail the original does not show. No
sharpening halos. Up to 2K.

What you'll get: Eyes that read as the focal point without a redrawn face.

Variation: For a group, name the person who is soft and say "match the sharpness of the others."

3. Remove Compression Artifacts

This JPEG has heavy compression: blocky patches in the [sky and
shadows] and ringing around edges. Smooth the blocks and remove the
ringing while keeping real edges and texture. Keep my face shape,
features, skin tone, expression, pose, colors, and framing exactly
as they are. Do not sharpen beyond the original, add detail, or
smooth skin. Output as a high-quality image up to 2K, same aspect
ratio.

What you'll get: Clean gradients and edges with the content untouched.

Variation: Add "the banding in the sky is the main problem" to focus the fix on one region.

4. Enhance a Screenshot of a Photo

This is a screenshot of a photo, so it is low resolution with
interface edges. Crop to the photo only, removing any bars, icons,
or borders. Then clean compression noise and restore clean edges
on hair, clothing, and the background. Keep every face shape,
feature, skin tone, expression, pose, and the original framing of
the photo exactly as they are. Do not invent detail. Up to 2K, same
aspect ratio as the cropped photo.

What you'll get: The photo without the interface, cleaner and still honest.

Variation: If you still have the original file, upload that instead; the screenshot is the weakest source.

5. Sharpen a Zoomed Crop

This is a tight crop from a larger photo, so it is soft and
pixelated. Enlarge it to [1:1 / 4:5] at up to 2K and bring hair
edges, clothing, and the background to clean sharpness. Keep my
face shape, eye shape and color, nose, mouth, skin tone, and
expression exactly as they are, soft rather than invented where
the crop shows no detail. No halos, no new texture, no smoothing.

What you'll get: A usable crop that stays soft where it should instead of guessing.

Variation: Brush the background only and ask for sharpening there, leaving the face as it is.

6. Fix a Low-Res Download

This photo was downloaded at a small size and shows soft edges and
blocky detail. Enlarge to up to 2K and restore clean edges on
hair, clothing, text, and the background, recovering detail only
where the original shows it. Keep every face shape, feature, skin
tone, expression, pose, and the framing exactly as they are. Do
not add texture, sharpen skin, or change colors. No halos.

What you'll get: A larger file with edges that hold and faces that match.

Variation: For a photo with text in it, add "spell any visible text exactly as it appears."

Low Light and Noise (7–11)

Noise reduction is where "enhance" quietly becomes "repaint," because the model cannot tell grain from skin texture unless you say which to keep. Describe the light that was actually in the room, say how much to brighten, and ask for texture to survive. A ChatGPT photo enhance prompt for night shots should keep the mood and make the faces readable, which is a different request from "make it bright."

7. Clean a Grainy Night Shot

This night photo is grainy and dark. Reduce the noise in the
[sky and shadows] by about half while keeping real texture in
skin, hair, and fabric, and lift the shadows just enough for my
face to read clearly. Keep the night mood, the warm street light
from the [right], and my face shape, features, skin tone,
expression, pose, and framing exactly as they are. No waxy skin,
no daylight look, no invented detail. Up to 2K.

What you'll get: A night photo that is still a night photo, with a readable face.

Variation: Add "keep the light trails and the neon as they are" for city scenes.

8. Fix Indoor Yellow Light

This indoor photo has a strong yellow cast from [ceiling lamps].
Neutralize the cast so whites read as white and skin looks the way
it does in daylight, then reduce the noise gently. Keep the warm
mood of the room at a lower strength. Keep my face shape, features,
natural skin tone, hair color, expression, pose, and framing
exactly as they are. Do not brighten the face beyond the room, do
not smooth skin. Up to 2K, same aspect ratio.

What you'll get: Natural skin and white walls with the room's warmth kept in check.

Variation: Name a neutral object, "use the white plate as the reference," to anchor the correction.

9. Recover a Backlit Silhouette

I am backlit by the [window / sunset] and my face is nearly a
silhouette. Lift the exposure on my face and body by about two
stops so my features read clearly, keeping the bright background
from blowing out and keeping the rim light on my hair. Recover
detail only where the original shows it. Keep my face shape,
eyes, nose, mouth, skin tone, expression, pose, and framing
exactly as they are. No HDR halo around my head, no invented
facial detail. Up to 2K.

What you'll get: A readable face in front of a window that still looks bright.

Variation: If the face is pure black in the original, lower expectations: ask for "a soft, dim recovery" instead of full exposure.

10. Soften Flash Harshness

This photo was shot with direct on-camera flash: flat bright face,
shiny forehead, a hard shadow on the wall behind me, and red-eye.
Soften the flash look: reduce the shine, lift the shadow on the
wall to a soft edge, fix the red-eye to my real eye color, and
restore gentle shading on my face as if the light came from a
softbox. Keep my face shape, features, skin tone, hair, expression,
pose, and framing exactly as they are. No smoothing, no new light
direction. Up to 2K.

What you'll get: A flash photo that reads like soft room light, same face.

Variation: Keep the flash look and ask only for "fix the red-eye and reduce forehead shine by half."

11. Fix Mixed White Balance

The [window] side of this photo is blue and the [lamp] side is
orange, so my face is two colors. Balance both sources so skin
reads as one natural tone across my face and whites read as
white, keeping a hint of warmth on the lamp side so the scene
still looks real. Keep the exposure, my face shape, features,
skin tone, hair, expression, pose, and framing exactly as they
are. No skin smoothing, no global warm filter. Up to 2K.

What you'll get: One skin color across the whole face, both light sources believable.

Variation: Brush the orange side only and ask it to "match the color of the other side."

HD and Upscale (12–16)

ChatGPT outputs up to 2K, so every "HD" and "4K" request lands at 2K; ask for "up to 2K" and you get what you asked for. For a pure resolution bump on a photo that is otherwise fine, a dedicated upscaler keeps the pixels more faithfully. ChatGPT is the right tool when the photo also needs light, color, or clean-up fixes in the same pass, and that is what these five are for.

12. Upscale to 2K Keeping Grain

Upscale this photo to 2K, same aspect ratio. Keep the film grain
and the slight softness of the original rather than replacing them
with a clean digital look. Sharpen only the edges that are already
sharp: hair, clothing seams, and the background. Keep my face
shape, features, skin tone, expression, pose, colors, and framing
exactly as they are. Do not invent detail, remove grain, or smooth
skin. No halos.

What you'll get: A bigger version of the same photo, grain intact.

Variation: Swap "keep the film grain" for "reduce the grain by a third" when the original is very noisy.

13. Prepare for Print

Prepare this photo for a [8 x 10 inch] print. Output at up to 2K
in a [4:5] ratio with my full head and shoulders in frame, clean
edges on hair and clothing, and shadows lifted slightly so they
hold on paper. Keep my face shape, features, skin tone, hair,
expression, pose, and colors exactly as they are, and keep real
skin texture. Do not oversharpen, add contrast, or invent detail.
No vignette, no border.

What you'll get: A print-ready file with shadows that will not block up on paper.

Variation: For a larger print, take the 2K result to a dedicated upscaler for the final size.

14. Enlarge an Old Small Scan

This is a small scan of an old print, soft and slightly faded.
Enlarge it to up to 2K, same aspect ratio, and restore clean edges
on clothing, the background, and any text, recovering detail only
where the scan shows it. Keep every face shape, feature, skin tone,
expression, pose, the framing, and the print's grain and era look
exactly as they are. Do not sharpen faces, modernize colors, or
invent details the scan does not hold. No halos.

What you'll get: A larger scan that still looks like the era it came from.

Variation: Rescan at the highest resolution first; a better source beats any prompt.

15. Enhance a Group Photo Where Every Face Must Stay

Enhance this group photo of [NUMBER] people: sharpen hair edges,
clothing, and the background, reduce noise gently, and lift the
shadows so every face reads clearly. Treat every face with the
same restraint: keep each person's face shape, eye shape, nose,
mouth, skin tone, hair, glasses, expression, and position exactly
as they are, and do not swap, blend, or redraw anyone. Recover
detail only where the original shows it. Up to 2K, same aspect
ratio. No smoothing, no halos.

What you'll get: A clearer group shot where everyone is still themselves.

Variation: Count the faces in the result against the original; if one drifted, brush that person's clothing and background only and rerun.

16. Enhance a Product Photo

Enhance this product photo for a listing. Sharpen the product's
edges and label text, clean noise from the background, and even out
the exposure so the [white / gray] backdrop is clean. Keep the
product's shape, proportions, colors, materials, label text,
logo, and reflections exactly as photographed; spell any text
exactly as it appears. Do not recolor, reshape, or add detail that
is not there. Up to 2K, 1:1, product centered.

What you'll get: A listing image with crisp edges and the label still accurate.

Variation: Add "replace the backdrop with pure white (#FFFFFF) and a soft contact shadow" if the background is uneven.

Color and Exposure (17–22)

Color fixes are global, so describe them and leave the brush alone. Say what the scene should look like afterward in concrete terms, because "warmer" turns faces orange and "pop" pushes saturation past natural. Name a neutral reference in the frame when you have one, and keep skin tone on the keep list by name.

17. Revive Faded Colors

The colors in this photo have faded to a flat, washed-out look.
Restore natural saturation and contrast: deeper blacks, clean
whites, and colors as they would have looked on the day, with
realistic skin. Keep my face shape, features, skin tone, hair,
expression, pose, and framing exactly as they are. Do not push
saturation past natural, add a filter look, or sharpen. Up to 2K,
same aspect ratio.

What you'll get: The photo as it looked before it faded, no filter on top.

Variation: Give known colors, "the car was dark green," to limit guesswork.

18. Recover an Overexposed Sky

The sky in this photo is blown out to white. Bring back a
believable [pale blue sky with soft clouds] that matches the time
of day and the light direction from the [upper left], blending
cleanly at the horizon and around hair and tree edges. Keep my
face, exposure on my face and body, pose, clothing, colors, and
framing exactly as they are. Do not darken the rest of the photo
or add drama. No halo at the skyline. Up to 2K.

What you'll get: A sky that belongs to the scene, with the subject untouched.

Variation: Brush the sky only so the rest of the frame is never regenerated.

19. Lift an Underexposed Subject

The background is correctly exposed but I am too dark. Lift the
exposure on me by about [1.5] stops, recovering detail in my
clothing and hair where the original shows it, and keep the
background at its current brightness. Keep my face shape,
features, skin tone, expression, pose, and framing exactly as
they are. Natural contrast, no HDR halos, no invented facial
detail, no noise smear. Up to 2K, same aspect ratio.

What you'll get: A subject that matches the background's exposure, same face.

Variation: Add "the shadows on my face will be noisy; reduce that noise gently without smoothing skin."

20. Keep Skin Tone Natural

Correct the exposure and color of this photo so my skin reads the
way it does in daylight: no orange, no gray, no pink shift. Use
the [white shirt / gray wall] as the neutral reference. Keep my
exact skin tone, face shape, features, hair, expression, pose,
and framing as they are, and keep real skin texture, freckles,
and marks. Do not lighten or darken my skin, smooth it, or add a
glow. Up to 2K.

What you'll get: Your actual skin tone under corrected light.

Variation: For a group with different skin tones, add "correct each person's skin to their own natural tone."

21. Vibrant Without Oversaturation

Make this photo more vibrant without oversaturating it: richer
[greens and blues] in the scene, cleaner whites, slightly deeper
contrast, and skin left natural. Protect skin tones and the
[red jacket] from oversaturation and keep gradients smooth with
no banding. Keep my face shape, features, skin tone, hair,
expression, pose, and framing exactly as they are. No filter
look, no vignette, no sharpening. Up to 2K, same aspect ratio.

What you'll get: Richer color that still looks like a photo.

Variation: Replace "richer" with "slightly muted, film-like" for the opposite direction.

22. Consistent Set

I have uploaded [NUMBER] photos from the same [event]. Match them
to the first photo: same exposure, white balance, contrast, and
color balance, so they read as one set. Correct only the
differences; do not restyle. In every photo keep each person's
face shape, features, skin tone, hair, expression, pose, and the
framing exactly as they are. No sharpening, no smoothing, no
crops. Output each at up to 2K in its original aspect ratio.

What you'll get: A matched set that looks like one camera and one afternoon.

Variation: On a paid plan, Thinking mode handles the whole batch in one request; on the free plan, run one photo at a time in the same chat.

Keep-the-Face Set (23–27)

These five are the strictest templates in the post, written for anyone searching for a ChatGPT prompt for photo editing without changing the face. Each one names features instead of saying "my face," limits the fix to named regions, and pairs with the selection brush: paint everything except the face, and the face is never regenerated. Fill in the feature placeholders from your own photo before you paste.

23. Sharpen Everything Except the Face

Sharpen this photo everywhere except my face. Bring the hair
edges, clothing, hands, and background to clean sharpness,
recovering detail only where the original shows it. Leave my face
pixel-for-pixel as it is: same face shape, eye spacing and color,
brow shape, nose width, lip shape, [MOLE / SCAR / DIMPLE], skin
tone, hairline, and expression. Do not touch, sharpen, or smooth
anything inside the face. No halos. Up to 2K, same aspect ratio.

What you'll get: A crisper photo with the face carried over unchanged.

Variation: Paint everything except the face with the selection brush before you send, so the face sits outside the regenerated region.

24. Brighten the Photo, Lock the Face

Brighten this photo by about [1] stop and lift the shadows, with
the change applied evenly so my face gets lighter only in exposure.
Lock my face: same face shape, eye shape and spacing, eye color,
brow shape, nose, lip shape, [MOLE / FRECKLES / SCAR], skin tone,
teeth, hairline, and expression as in the upload. No smoothing,
no reshaping, no whitening, no catchlights added. Keep pose,
clothing, and framing exactly as they are. Up to 2K.

What you'll get: The same photo, one stop brighter, nothing else moved.

Variation: Brush the background and clothing for the brightening and leave the face unpainted, then ask for "match the face exposure to the rest" in a second pass only if it looks dark.

25. Denoise With Features Named

Reduce the noise in this photo by about half, in the background,
clothing, and shadows first, and leave skin texture alone. Lock
my face: same face shape, eye shape, spacing, and color, brow
shape, nose width, lip shape, [MOLE / FRECKLES / SCAR], skin
tone, pores, hairline, and expression as in the upload. Do not
smooth, reshape, or redraw any part of the face. Keep pose,
framing, and colors exactly as they are. Up to 2K, same aspect
ratio.

What you'll get: Less grain in the background, real skin on the face.

Variation: Brush only the background and shadows, so the denoise never reaches the face at all.

26. Upscale With a Feature Checklist

Upscale this photo to up to 2K, same aspect ratio, as a faithful
enlargement: clean edges on hair, clothing, and background,
recovering detail only where the original shows it. Checklist for
my face, every item unchanged: face shape, eye shape and spacing,
eye color, brow shape, nose, lip shape, teeth, [MOLE / FRECKLES /
SCAR], skin tone, hairline, and expression. Soft where the
original is soft; do not invent facial detail. No halos.

What you'll get: A larger file where the face is soft in the same places it was before.

Variation: An upscale touches every pixel, so the brush cannot exclude the face here; if the face still drifts, use a dedicated upscaler for this photo.

27. Full Enhance With a Face Lock

Enhance this photo in one pass: correct exposure and white
balance, reduce noise by half, sharpen hair edges, clothing, and
background, and output at up to 2K, same aspect ratio. The face is
locked: same face shape, eye shape, spacing, and color, brow
shape, nose width, lip shape, teeth, [MOLE / FRECKLES / SCAR],
skin tone, pores, hairline, and expression as in the upload,
changed only by the global exposure and color correction. No
smoothing, reshaping, or invented detail anywhere on the face.

What you'll get: Every fix at once, with the face held to a written checklist.

Variation: Split it in two: brush everything except the face for the sharpening and denoise, then run the exposure and color correction unbrushed with the same lock.

Old Phone Photos and Scans (28–30)

Old sources carry damage the model has never seen in your current photos: early sensor noise, a purple cast, scratches, and the compression a messaging app adds on every forward. Name the source so the fix matches the damage, and forbid modernizing the look. These are also the photos most likely to be of other people, so check consent before you upload.

28. Early Smartphone Photo

This photo is from an early smartphone: [2 to 5] megapixels, soft,
noisy, with a slightly purple cast and smeared shadows. Enlarge to
up to 2K, same aspect ratio, clean the noise while keeping fabric
and hair texture, neutralize the cast, and restore clean edges
where the original shows them. Keep every face shape, feature,
skin tone, expression, pose, and the framing exactly as they are.
Do not invent facial detail or modernize the look. No halos.

What you'll get: A cleaner, larger version that still looks like the year it was taken.

Variation: Add "keep the slight softness; it is part of the photo" if the result looks too clinical.

29. Scanned Print With Scratches

This is a scan of a printed photo with scratches, dust spots, and
a [crease across the lower left]. Remove the damage by filling it
with the matching surrounding tones and grain, then enlarge to up
to 2K, same aspect ratio. Keep every face shape, feature, skin
tone, expression, pose, the framing, the colors, and the print's
grain exactly as they are. Do not sharpen faces, smooth skin,
colorize, or reconstruct anything the damage fully covers.

What you'll get: A clean print at a usable size, faces and grain intact.

Variation: Brush the crease first and run only the damage line, then enlarge in a second pass.

30. WhatsApp-Compressed Photo

This photo was sent through a messaging app and compressed hard:
small size, blocky shadows, smeared hair, and ringing at the
edges. Enlarge to up to 2K, same aspect ratio, smooth the
compression blocks, remove the ringing, and restore clean edges
on hair, clothing, and the background, recovering detail only
where the original shows it. Keep every face shape, feature, skin
tone, expression, pose, and framing exactly as they are. No
invented detail, no skin smoothing, no halos.

What you'll get: The photo as it looked before the app squeezed it, as far as the file allows.

Variation: Ask the sender for the original as a document attachment; most apps skip compression that way, and a better source beats any prompt.

Tips for Better ChatGPT Photo Enhance Prompt Results

Upload the best file you have. The model can only keep what it can see. The original from the camera roll beats the copy from a chat thread, and the chat copy beats a screenshot. If the only version left is compressed, say so in the prompt so the fix targets compression instead of blur.

Name the problem, then the permission. "Enhance" means nothing to the model; "motion blur, left to right" or "underexposed by about two stops" does. Then say what may be sharpened and what may not. The prompts above spend as many words on permission as on the fix, because that is where the face is protected.

Treat "4K" as "up to 2K." ChatGPT outputs up to 2K resolution. Asking for 4K or 8K gets you a 2K file, so the number buys nothing. If you need a larger print, enhance in ChatGPT first, then run the result through a dedicated upscaler for the final size.

Check the face at 100% before you accept. Compare eyes, teeth, hairline, and any mole or scar against the original side by side. Thinking mode on paid plans runs its own verification pass, and it checks that the face is plausible, which is a lower bar than yours. If something drifted, restart from the original file; a second edit on a drifted result drifts further.

Pick the right tool for the job. For a pure resolution bump on a photo that needs nothing else, a dedicated upscaler keeps pixels more faithfully. For a hard local edit where nothing else may move, Gemini's Nano Banana is stronger, and the 50 best Nano Banana prompts collection covers that side. ChatGPT is the right choice when the photo needs light, color, and clean-up fixed together.

Generate Custom ChatGPT Photo Enhance Prompts

Your photo has its own damage, light, and face to protect. Describe the problem in plain English to the image prompt generator and it writes the full enhance instruction, keep list included; the Template Builder saves a feature checklist you can paste into every prompt for the same person.

For edits beyond enhancement, the hub post 40 ChatGPT photo editing prompts covers backgrounds, restoration, and restyles. If you came looking for ChatGPT photo editing commands, that post explains why there are none and what to type instead, and ChatGPT prompts for pictures of yourself covers the edits that change the scene while keeping you in it.

Name the blur, lock the face, ask for up to 2K, and check at 100%.