A quick note before this article. I put together a guide covering 500+ tested AI image styles and prompts for Midjourney, DALL-E, Stable Diffusion, Leonardo, and more. Every prompt is copy-paste ready and organized by category: cinematic styles, photography looks, fine art, 3D and CGI, fantasy, and dozens more. If photo editing prompts are just the start for you, the full toolkit is right here. Now, back to the article.
You spent 20 minutes writing the perfect description. The AI gave you a photo where the background looks right, the lighting is wrong, your subject's face looks slightly melted, and the outfit you specifically said to keep the same is now completely different.
That is not a tool problem. That is a prompt problem.
Every bad AI photo editing prompt fails the same way: it tells the AI what to change but gives it no instructions on what to protect. Fix that one thing and the results shift completely.
Here is the formula, followed by 40 copy-paste AI photo editing prompts organized by edit type.
The Four-Part Formula That Fixes Everything
The best AI photo editing prompts always include four things.
Action verb. What should the AI do? Replace, remove, enhance, restyle, brighten. Be direct and use one verb per prompt.
Target element. What exactly is changing? Not "the photo" but "the background behind the subject" or "the skin under the eyes and around the nose."
Desired result with a real reference. Vague descriptions produce vague outputs. "Nice lighting" tells the AI nothing. "Warm golden-hour light from the upper left, soft and directional" is something it can actually render. Name a film stock. A decade. A director. A specific color palette. The more concrete the reference, the less the AI guesses.
Lock line. This is the most important part, and most people skip it entirely.
The lock line tells the AI what it is not allowed to touch. "Keep the subject's face, pose, and clothing identical" is a lock line. Without it, the AI treats the entire image as editable and will "improve" things you never asked it to change.
A complete prompt using all four parts looks like this:
Replace the background with a golden-hour beach scene, warm horizon light from behind the subject, soft bokeh, and a long shadow stretching toward the camera. Match the rim light on the subject to the new scene direction. Keep the subject's face, pose, and outfit entirely untouched.
That prompt names the action, the target, a specific visual reference, and a clear lock. It leaves the AI very little room to guess.
Why Your AI Edits Keep Looking Fake
Here's the part most guides skip entirely.
AI photo edits look fake for three specific reasons. First, you changed the background but did not describe the lighting direction. A person standing in a new scene looks composited rather than natural because the light hitting their face no longer matches the environment behind them. The fix is to describe where the new light comes from and tell the AI to match it to the subject's existing shadows.
Second, you stacked too many edits into one prompt. "Fix exposure, retouch skin, change background, add color grade" in a single sentence gives the AI four competing instructions and the freedom to prioritize whichever one it wants. It will prioritize wrong. Run one edit per prompt, in sequence, building on the previous result.
Third, you used emotional language instead of visual language. "Make it cinematic" means nothing to an image model. "Kodak Portra 400, warm skin tones, slightly faded blacks, subtle grain" is something it can actually render. Every mood word you use is a guess the AI has to make on your behalf.
Now here are the prompts.
Background Swap Prompts
Change the scene, lock the person. Describe the new lighting so the subject blends naturally instead of looking pasted in.
1.
Replace the background with a clean studio backdrop, neutral warm gray, three-point lighting from the front. Keep the subject's face, pose, outfit, and skin tone exactly as they appear. No changes to the subject.
2.
Replace the background with a rain-slicked Tokyo street at night, neon signs reflected in the puddles, soft depth of field behind the subject. Match the cool ambient light color to the subject's existing skin highlights. Keep the pose and outfit untouched.
3.
Replace the background with a minimalist white loft interior, large window with natural diffused daylight from the left. Do not change the subject's face, hair, skin tone, or clothing in any way.
4.
Swap the background for an overcast mountain trail, cool and diffused light, fog in the distance. Reflect the cool light temperature subtly on the subject's shoulder facing the scene. Keep all facial features, expression, and clothing identical.
Portrait Retouching Prompts
Retouching fails when models over-smooth until the subject looks plastic. The phrase "keep skin texture natural" is your lock line for every portrait edit.
5.
Enhance the skin in this portrait: remove visible blemishes, soften the under-eye shadows, and even the tone slightly. Keep natural pores and skin texture intact. Do not alter facial structure, hair, or any lighting.
6.
Brighten the teeth naturally. Remove the yellow cast. Keep the natural tooth shape and size. No whitening level that looks cosmetic or artificial.
7.
Reduce redness around the nose and cheeks. Keep all other skin tones, facial features, and lighting completely unchanged. Subtle correction only.
8.
Editorial face polish: even the skin, add a subtle contour, boost the lips slightly, add a touch of warmth to the highlights. Keep the facial features, hair, and existing lighting identical.
Style Transfer Prompts
Name a specific visual reference or the AI will invent one. Invented references are where faces start to fall apart.
9.
Apply Kodak Portra 400 film aesthetic: warm skin tones, subtle grain, faded blacks, muted saturation throughout. Keep the subject's face and pose completely identical.
10.
Restyle with a Wes Anderson visual look: symmetrical composition, pastel color blocks (mustard, dusty pink, mint green), flat front lighting, retro framing. Keep the subject's face and pose unchanged.
11.
Apply a 1970s vintage magazine editorial look: warm overexposed tones, subtle vignette, period film grain, slightly faded contrast. Keep the subject's face, body proportions, and composition identical.
12.
Restyle as an oil painting in an impressionist palette, visible brushstrokes, soft edges, painterly textures. Keep the subject recognizable and the composition unchanged. Do not alter facial proportions.
Outfit Change Prompts
Write outfit prompts like a stylist would: material, fit, color, detail. Then lock everything else with the same precision.
13.
Replace the outfit with a tailored charcoal wool suit, slim cut, crisp white shirt, slim leather belt, polished black oxfords. Professional fit. Keep the face, hairstyle, body proportions, and background completely identical.
14.
Replace the outfit with an oversized cream linen shirt, relaxed dark-wash jeans, and clean white trainers. Effortless casual look. Keep the face, hair, and background untouched.
15.
Replace the outfit with a floor-length silk gown in deep emerald, subtle shimmer, clean drape. Keep facial features, body proportions, pose, and background identical.
16.
Replace the jacket only with a vintage brown leather bomber, worn collar, brass zip, slightly distressed. Keep all other clothing items, face, hair, and background exactly as they are.
Color Grading Prompts
Think in film references, not in moods. Moods are subjective. Film references are something the model can actually reproduce.
17.
Apply a warm cinematic grade: amber highlights, rich shadows, film-like contrast, slightly faded blacks. Do not change the framing, subject, or facial expression.
18.
Apply a teal and orange blockbuster grade: cool teal in the shadows, warm orange skin tones, high contrast, deep blacks. Keep the composition and subject completely identical.
19.
Apply a muted Scandinavian editorial grade: desaturated tones, cool midtones, lifted blacks, clean and flat look. Do not alter the subject or composition.
20.
Apply a neon cyberpunk grade: high saturation, magenta and cyan in the highlights, deep crushed blacks. Keep the composition and subject unchanged.
Object Removal Prompts
The most important part of an object removal prompt is describing what fills the space behind the removed object. Without that, the AI fills it randomly.
21.
Remove the bystander on the right edge of the frame. Reconstruct the wall behind them, continuing the brick pattern and shadow naturally. No visible blending seam or distortion in the rebuilt area.
22.
Remove the coffee cup from the desk in the lower right corner. Fill the space with a natural continuation of the wood grain surface. Match the lighting and texture of the surrounding table area.
23.
Remove the logo on the subject's shirt. Reconstruct the fabric texture, pattern, and natural fold of the garment where the logo appeared. Keep the rest of the shirt and the full subject untouched.
24.
Remove the power lines crossing the upper third of the sky. Fill with a natural continuation of the sky tone, color gradient, and cloud coverage. Keep the buildings, trees, and foreground completely untouched.
Old Photo Restoration Prompts
Restore the damage. Do not modernize the photo. The era is part of the image and the repair should respect it.
25.
Repair physical damage on this vintage photograph: remove visible scratches, creases, and tears. Rebuild missing areas naturally. Preserve the original grain, lighting, and period aesthetic throughout.
26.
Colorize this black-and-white photo with period-accurate colors. Keep the grain intact. Use a muted 1950s palette with warm shadows. Do not over-saturate or apply any modern color treatment.
27.
Recover the faded color in this old photograph: restore lost contrast, rebuild the original color range, preserve the softness and warmth of the era. Do not apply modern color grading.
28.
Sharpen and restore detail in this blurry vintage portrait. Reconstruct facial features using natural proportions from visible reference points in the image. Keep skin tone and lighting unchanged. Preserve the period feel.
Product Photography Prompts
Specify the platform. Amazon wants pure white. Instagram wants lifestyle context. Pinterest wants flat lay. Each one expects a different look.
29.
Place this product on a pure white background for an Amazon listing. Remove all shadows except a clean soft drop shadow directly beneath the product. Keep every product detail, label text, and color exactly as-is.
30.
Place this product on a weathered oak table in a warm kitchen setting. Natural window light from the left. Lifestyle editorial feel for Instagram. Keep all product labels and packaging details fully legible.
31.
Place this product in a flat-lay composition on white marble with a few scattered coffee beans nearby. Overhead angle. Clean daylight. Suitable for Pinterest. Keep all product packaging detail and original colors exactly as they appear.
32.
Remove the existing background and place the product centered on a bold primary red background. Clean edges around the product. Keep all label text, product detail, and original color fully intact.
Creative Effects Prompts
Creative prompts need one visual reference and one lock. Use both or the AI will treat the entire image as a canvas.
33.
Add a double exposure effect: blend this portrait with a dense forest landscape so the trees and light filtering through leaves are visible through the subject's face and shoulders. Keep the subject's outline clear and the face readable at full composition.
34.
Add a glitch art effect with horizontal color shift artifacts, mild pixel distortion on the left edge, and slight chromatic aberration throughout. Keep the subject's face central and fully readable.
35.
Add falling cherry blossom petals in front of the subject, rendered realistically at varying depths with some slightly blurred in the foreground. Do not change the subject, background, or existing lighting.
36.
Apply a soft light leak from the upper right corner, warm amber and pale pink, organic film feel. Keep all subject details, composition, and color temperature of the scene unchanged.
Lighting Fix Prompts
Describe the light source first. Then tell the AI what to preserve. Order matters here.
37.
Add a warm golden-hour rim light from the upper left, soft and natural. Blend it realistically with the existing ambient light on the subject. Do not alter facial features, expression, or background in any way.
38.
Brighten the midtones by approximately 20% and add warmth to the highlights. Keep shadow detail fully intact throughout. Do not alter the background or the subject's face.
39.
Correct the harsh overhead flash in this indoor portrait: soften shadows under the eyes and nose, balance skin tones across the face, add slight warmth to the overall tone. Keep facial features completely identical.
40.
Add a soft window light from camera left: diffused, natural, cool white. Blend the shadow side of the face to match the new source naturally. Do not change the subject's pose, expression, or outfit.
Which AI Tool Actually Handles Which Edit
This is the question most articles never answer directly.
ChatGPT (GPT-4o) handles natural language well and is the best starting point for most people. It works well for background swaps, color grading, basic retouching, and creative effects. If you are not sure where to start, start here.
Gemini tends to follow lock lines more consistently than other models in direct testing. It preserves the existing photo more faithfully, which makes it reliable for restoration, subtle retouching, and edits where accuracy to the original matters most.
Midjourney is built for stylized outputs. Use it for style transfers, color grading, and artistic looks. It is not designed for precise object removal or surgical portrait editing. Use it for look and feel, not for controlled changes.
Adobe Firefly (Generative Fill) is the strongest tool for object removal and background replacement on real photographs. It works on selections rather than full-image prompts, which gives you far more control over what changes and what does not. If photographic realism is the priority, this is the right tool.
Picsart AI is the simplest option for mobile editing. Keep your prompts shorter and more direct when using it, since the model responds better to concise instructions than detailed multi-part prompts.
Three Mistakes That Ruin Technically Good Prompts
You can follow the four-part formula exactly and still get bad results if you fall into one of these.
Stacking multiple edits in one prompt. Run one change per prompt. If you want a background swap followed by a color grade, do them as two separate passes in sequence.
Using mood language instead of visual language. "Feel magical" or "look dramatic" are not instructions. "Add soft diffused light, warm golden tones, and shallow depth of field in the background" is an instruction. Every mood word is a guess you are handing to the AI.
Not specifying which part of the image is being changed. "Fix the lighting" is not a prompt. "Brighten the midtones on the subject's face while keeping shadow detail in the hair and background intact" tells the AI exactly where to work and what to leave alone.
Frequently Asked Questions
Can I use these prompts with the free version of ChatGPT? Yes. Image editing features are available through the ChatGPT interface when you upload a photo, and most of these prompts work on the free tier without any changes.
Why does the AI keep changing my subject's face even when I tell it not to? Your lock line is probably not specific enough. Instead of "keep the face the same," try: "Keep the facial structure, skin tone, eye shape, eyebrow shape, and expression completely identical. Do not reconstruct, smooth, or alter any facial feature."
What is the difference between AI image generation and AI photo editing? Generation creates an image from nothing using a text description. Photo editing takes an existing image and modifies specific parts of it. The prompts in this guide are for editing photos you already have, not generating new ones from scratch.
Do these prompts work on Midjourney? The style transfer, color grading, and creative effects prompts work well on Midjourney. For object removal and portrait retouching, use Photoshop Generative Fill or ChatGPT instead. Midjourney is not optimized for surgical photo editing.
Are AI-edited photos safe for commercial use? It depends entirely on the tool. ChatGPT, Adobe Firefly, and Picsart have commercial-friendly terms for their edited outputs. Midjourney's commercial rights depend on your subscription tier. Always check the specific tool's terms before using edited images in paid work.
The prompt is rarely the problem. The missing lock line is the problem.
Every AI photo edit that looks fake, over-processed, or artificially altered can be traced back to a prompt that told the AI what to change but gave it no instructions on what to leave alone. Add the lock line. Name a real visual reference instead of a mood word. Run one edit at a time.
That is the whole shift.
Which edit type has been giving you the most trouble? Drop it in the comments and I'll write a specific prompt for it.
