AI Photo Effect Prompts: 4 Fashion Portrait Effects You Can Make in ChatGPT

By Clayton H · Published 7 November 2026 · Tools: ChatGPT

AI photo effect prompts turn a plain portrait into an art-directed image by combining it with a second image that holds only the effect: a pane of reeded glass, a cloud of soap bubbles, a sheet of organza or a burst of pigment dust. You generate the portrait, generate the effect on its own, then upload both to ChatGPT with a combining prompt that keeps the person exactly as they are and adds the effect around them.

This guide gives you all four effects as copy-ready prompts with blanks for your own subject, colors and framing. It takes three prompts per effect and works with ChatGPT image generation, which is available on every ChatGPT plan, including free.

What are AI photo effect prompts?

An AI photo effect prompt is an instruction that adds one physical effect, such as glass, bubbles, fabric or dust, to an existing photo while leaving the subject unchanged. The method here splits that into two images, because a model reproduces an effect far more reliably when it can see it on its own first than when it has to invent it from a description.

What you need

• ChatGPT: for all three steps. The combining step needs a tool that accepts two images in one message, and ChatGPT does.
• A portrait: one you generate with the prompt below, or your own photo. Plain background, clean light and one strong color work best.
• Three generations per effect: the portrait, the effect image and the combine, plus a retry if the combine drifts.

How to add an AI effect to a portrait in ChatGPT

Every effect uses the same three steps. The prompts for each effect further down already have the details filled in, so read this section once and then jump to the effect you want.

Step 1: Generate the portrait

Generate a fashion portrait with one strong idea in it: a bold accent color, a sculptural piece of clothing or a pose with a clear gesture. Keep the background plain and the light clean. If you are using your own character, attach their reference image and start the prompt with: “Use the attached image as the identity reference only. Match the face and hair exactly as shown. Restyle the wardrobe as described below.”

A high-fashion editorial portrait [SHOT TYPE, e.g. from the waist up] of [SUBJECT, e.g. a woman in her twenties]. She wears [OUTFIT, e.g. an oversized dove gray wool blazer with long oxblood patent opera gloves]. [POSE, e.g. one gloved hand gripping the lapel, chin lifted, a cool level gaze into the lens]. [MAKEUP, e.g. bare freckled skin and a matte oxblood lip]. A seamless [BACKGROUND COLOR + HEX, e.g. cool gray #B9BEC1] backdrop, [LIGHT, e.g. soft directional window light from the left]. Shot on medium format, [LENS, e.g. 80mm lens], crisp detail, subtle film grain. [ASPECT RATIO, e.g. square 1:1] format, photorealistic.

Step 2: Generate the effect on its own

Generate the effect with no person in it, on a plain background, in the same aspect ratio as the portrait. Each effect below has its own effect prompt.

Step 3: Combine the two images

Open a new ChatGPT message, upload the portrait first and the effect image second, then paste the combining prompt. The order matters, because the prompt refers to “the first image” and “the second image”.

Use the first image as the subject and the second image as the effect. Keep the first image exactly as it is: the same pose, face, framing and colors. Do not redraw or re-pose the subject. [ACTION, e.g. Surround the subject with the floating bubbles from the second image]. Keep the subject’s face fully readable. Keep the [ELEMENT, e.g. size, sheen and spacing of the bubbles] from the second image. The result should look like one photograph, not a collage. [ASPECT RATIO] format, photorealistic, high detail.

How to make the reeded glass effect with AI

Half of the portrait sits behind a pane of fluted glass, so the face and clothes break into repeating vertical slivers.

Portrait tip: give the outfit one strong accent color, such as red gloves or a bright lip. The accent is what makes the slivers stand out.

Effect prompt:

A single pane of clear reeded glass seen straight on, filling the frame: tall, evenly spaced vertical fluted ribs, each one catching a thin bright highlight along its edge, set against a plain flat [BACKGROUND COLOR + HEX, e.g. cool gray #B9BEC1] background that shows softly distorted through the ribs. No people, no objects, no text. Flat to the camera, [ASPECT RATIO] format, high detail, no vignette.

Combining prompt:

Use the first image as the subject and the second image as the glass. Keep the first image exactly as it is: the same pose, face, framing and colors. Do not redraw or re-pose the subject. Place the reeded glass pane from the second image in front of [SIDE, e.g. the right half] of the subject, with its edge running as a clean vertical line just past the bridge of the nose. The uncovered side stays perfectly sharp. Behind the glass, everything refracts through the vertical ribs, so the face, the clothes and [ACCENT, e.g. the oxblood gloves] break into repeating offset vertical slivers. Keep one eye fully sharp and readable. Keep the rib width and highlights from the second image. The result should look like one photograph shot through a real sheet of fluted glass, not a collage. [ASPECT RATIO] format, photorealistic, high detail.

How to make the floating bubbles effect with AI

The subject is surrounded by iridescent soap bubbles, some drifting in front and bending the image behind them, some floating behind.

Portrait tip: this works best outdoors against open sky or on a plain colored backdrop. Movement in the pose, such as a raised arm or fabric caught by wind, makes the bubbles feel natural.

Effect prompt:

Dozens of clear iridescent soap bubbles floating across a plain flat [BACKGROUND COLOR + HEX, e.g. cobalt blue #1F5FBF] background, in a mix of sizes from tiny to very large, scattered loosely with uneven spacing, some overlapping and some cut off by the edges of the frame. Each bubble is thin and transparent, with a soft rainbow sheen in pink, amber and violet and crisp white highlights. No people, no text, no other objects. Flat to the camera, [ASPECT RATIO] format, high detail.

Combining prompt:

Use the first image as the subject and the second image as the bubbles. Keep the first image exactly as it is: the same pose, face, framing and colors. Do not redraw or re-pose the subject. Surround the subject with the floating bubbles from the second image, keeping their sizes, sheen and loose spacing. Some bubbles float in front of [FRONT AREA, e.g. the dress and arm], and some drift behind the subject against the background. The bubbles in front slightly bend and magnify what is behind them, and all of them catch the same light as the photograph. Keep the face completely clear of bubbles. The result should look like one photograph taken in a cloud of real bubbles, not a collage. [ASPECT RATIO] format, photorealistic, high detail.

How to make the organza veil effect with AI

A sheet of sheer colored organza blows across the lower half of the face, with the features showing through and the loose end trailing in the wind.

Portrait tip: use a head-and-shoulders shot or a close-up, so the face is large enough for the veil to read. Choose an organza color that stands out against the outfit, such as tangerine on teal.

Effect prompt:

A single large sheet of sheer [FABRIC COLOR + HEX, e.g. tangerine #FF8A3D] organza fabric floating and twisting in mid-air on a plain flat [BACKGROUND COLOR + HEX, e.g. teal #3E8C8C] background, caught mid-flutter with soft folds and rippling edges, translucent enough that the background shows through it, with a fine fabric sheen where it catches the light. No people, no text, no other objects. Flat to the camera, [ASPECT RATIO] format, high detail.

Combining prompt:

Use the first image as the subject and the second image as the fabric. Keep the first image exactly as it is: the same pose, head angle, face, framing and colors. Do not redraw or re-pose the subject. Wrap the sheer [FABRIC COLOR] organza from the second image across the lower half of the subject’s face, from just below the eyes down to the chin, tucking loosely into [NECKLINE, e.g. the collar of the coat], which stays clearly visible. A loose end of the organza trails out behind the head into the background, as if caught by wind. The lips and jaw show through the translucent fabric, softened and tinted [FABRIC COLOR], and the eyes stay completely clear and sharp above it. Keep the folds, sheen and translucency from the second image, lit by the same light as the photograph. The result should look like one fashion photograph, not a collage. [ASPECT RATIO] format, photorealistic, high detail.

How to make the pigment dissolve effect with AI

One side of the subject breaks apart into fine colored dust that drifts away across the background, while the face stays intact.

Portrait tip: a strongly colored garment works best, because the dust takes its color from the clothes. Turn the body slightly so the dissolving side has room to drift into.

Effect prompt:

A cloud of fine [DUST COLOR + HEX, e.g. crimson #B3122E] pigment dust drifting from left to right across a plain [BACKGROUND COLOR + HEX, e.g. warm putty #CDBFA8] background: dense and almost solid along the left edge, then breaking into thousands of tiny particles and soft powdery wisps that thin out and scatter toward the right of the frame. No people, no text, no objects. Flat to the camera, [ASPECT RATIO] format, high detail, no vignette.

Combining prompt:

Use the first image as the subject and the second image as the pigment. Keep the first image exactly as it is: the same pose, face, framing and colors. Do not redraw or re-pose the subject. Make [SIDE, e.g. the right side] of the subject disintegrate into the pigment dust from the second image: [PARTS, e.g. the shoulder, the back of the jacket and the outer edge of the collar] break apart into fine particles that drift away across the background, following the direction and density of the dust in the second image. The face, eyes and lips stay completely intact and sharp. Keep the styling, colors, light and background from the first image. The result should look like one photograph of a figure turning to dust in real time, not a collage. [ASPECT RATIO] format, photorealistic, high detail.

Things worth knowing

• Match the aspect ratios. If the portrait and the effect image differ, the model crops one of them and the effect lands off center.
• Keep the effect image to one element on a plain background. A busy effect image is behind most failed results, so when something goes wrong, regenerate that first.
• Keep the upload order. Portrait first, effect second, and keep the words “the first image” and “the second image” in the prompt. Vague references are the fastest way to get the subject and the effect mixed up.
• Expect the occasional redraw. If ChatGPT gives the person a new pose or face, run the combining step again. The line “Keep the first image exactly as it is” is there to prevent this, so do not remove it.
• Take the effect color from the portrait. Crimson dust from a crimson jacket, a cobalt bubble background from a cobalt sky. That match is what makes the result look designed rather than stacked.
• Rescue a hidden face with one line. If the face disappears into the effect, add “keep the face fully readable and sharp” and run the combining step again rather than starting over.

FAQ

Can I use these prompts on my own photos?

Yes. Skip Step 1, upload your own portrait as the first image, and fill in the blanks in the combining prompt to match it. Plain backgrounds and clear light give the cleanest results.

Do I need a paid ChatGPT plan?

No. OpenAI says ChatGPT Images is available on all ChatGPT tiers, including free.

Why not describe the effect in one prompt instead of making a separate image?

Because the model copies an effect it can see far more consistently than one it has to imagine. The separate effect image fixes the look of the glass, bubbles, fabric or dust before the person is involved, so the combine only has to place it.

Why did ChatGPT change the face or pose?

The combining step regenerates the whole image, so some drift can happen on any run. Keep the “Keep the first image exactly as it is” line, keep the upload order, and run it again; a second attempt usually holds the pose.

Does this work in other AI image generators?

Usually, if the tool accepts two reference images in one prompt. The prompts are written in plain language with no ChatGPT-only syntax, but results vary by model, so expect to adjust the blanks.

Which effect is easiest to start with?

Floating bubbles. It is the most forgiving, because the bubbles sit around the subject rather than across the face, so small placement errors still look natural.

Sources

• Images in ChatGPT, OpenAI Help Center: image generation and editing in ChatGPT, uploading images, aspect ratios and plan availability.