---
title: "ChatGPT Photo Prompts: 6 Templates for Edits and Feedback"
description: "Copy-ready ChatGPT photo prompts for editing, critique, alt text and product listings. Each one says what to keep, what to change and what you want back."
url: https://www.promptpal.app/blog/chatgpt-photo-prompts-templates
author: "Julius Olajumoke"
published: 2026-11-08T08:04:15.700Z
updated: 2026-11-18T22:20:18.670Z
category: "General"
tags: ["chatgpt", "photo prompts", "image editing", "prompt templates", "prompt engineering"]
language: en
reading_time_minutes: 8
image: https://images.ctfassets.net/xzkuxcljsk12/iohs9pEhsxmZNf7vDQP1w/c2fb27b36faea5725d4a0f70b5aa6b43/image.png
publisher: PromptPal
---

# ChatGPT Photo Prompts: 6 Templates for Edits and Feedback

Copy-ready ChatGPT photo prompts for editing, critique, alt text and product listings. Each one says what to keep, what to change and what you want back.

The best ChatGPT photo prompts say what the photo is for, what must stay untouched, and what you want back. Upload the image, state those three things, and you get usable edits, honest critique or finished copy instead of a vague "nice photo!"

Below are six prompts you can paste today, plus a worked example showing why the vague version fails. Image features vary by ChatGPT plan and model version, so check what your version can do. Some can only look at a photo and describe or advise. Others can also return an edited image.

## Start by deciding which kind of photo prompt you need

Most people type "make this photo better" and get a shrug in paragraph form. The problem is that "better" could mean three different jobs.

- **Edit the image.** Change the background, remove clutter, adjust lighting. This only works if your ChatGPT version returns edited images.
- **Get advice on the image.** Critique the composition, suggest crop and colour changes you then make in Lightroom, Photoshop, or your phone.
- **Turn the image into text.** Alt text, captions, product descriptions, a listing, a report note.

Pick one per prompt. If you ask for all three at once, each gets a thin answer. Run them as separate messages in the same chat, since the photo stays in context.

## The four-part structure that works for photo prompts

Every prompt below follows the same skeleton. It is the same idea as in our guide to [writing better ChatGPT prompts](https://promptpal.app/blog/how-to-write-better-chatgpt-prompts-a-complete-guide-2026), applied to images.

1. **Purpose.** Where will the photo end up? An Etsy listing, a LinkedIn profile, a slide, a printed menu.
2. **Keep.** What must not change. Colours of the product, a person's face, the text on a label.
3. **Change.** The specific things you want different, listed one by one.
4. **Output.** The format you want back: a numbered list of edits, one edited image, a 20-word caption.

The "keep" line is the one people skip, and it causes most of the disappointment. When you don't say what's fixed, the model treats everything as negotiable. A blue mug comes back teal. A logo gets redrawn.

## Six copy-ready ChatGPT prompts for photo editing and feedback

Replace the `[placeholders]`, upload the photo, and send.

### 1. Clean up a product photo

Use this for listings, catalogues and social posts where the product has to look like the real thing.

> `I'm uploading a product photo for [a marketplace listing / my website / an Instagram post]. The product is [product name and one-line description].`
>
>
>
> `Keep exactly as they are: the product's shape, colour, texture, any text or logo on it, and its proportions.`
>
>
>
> `Change only this:`
>
> `- Replace the background with [plain white / soft grey / a light wooden surface].`
>
> `- Remove [the cables, the shadow on the left, the clutter behind it].`
>
> `- Even out the lighting so no part of the product is blown out or too dark.`
>
>
>
> `If you can return an edited image, do that. If you can't, give me a numbered list of the exact edits to make in a photo editor, with approximate slider directions (for example, exposure up slightly, highlights down).`

The fallback sentence at the end matters. It means you get something useful whether or not your version produces images.

### 2. Swap or simplify a headshot background

> `This is a headshot for [a LinkedIn profile / a company team page]. Keep the person's face, hair, skin tone, expression and clothing unchanged. Do not retouch or reshape anything about their appearance.`
>
>
>
> `Replace the background with [a neutral light grey / a softly blurred office]. Keep the edges around hair and shoulders natural.`
>
>
>
> `If you can't edit the image, tell me what is distracting in the current background and how to reshoot or crop it.`

The line about not retouching is deliberate. You want a background change, not a different-looking person.

### 3. Critique composition and lighting

This is the prompt to use when you want a second pair of eyes and plan to do the edits yourself.

> `Act as a photo editor giving feedback on this image. It was taken with [phone model / camera and lens] and will be used for [purpose].`
>
>
>
> `Give me:`
>
> `1. The single biggest weakness, in one sentence.`
>
> `2. Three specific fixes, ordered by impact. For each, say whether it's a reshoot fix or an editing fix.`
>
> `3. A suggested crop, described in plain terms (for example, crop in from the right until the subject sits on the left third).`
>
> `4. One thing that already works and I should keep.`
>
>
>
> `Be direct. Don't pad with praise.`

Asking for "the single biggest weakness" forces a ranking. Without it you get a flat list of five equally weighted tips.

### 4. Write alt text

> `Write alt text for this image. It appears on [a blog post about topic / a product page for product].`
>
>
>
> `Rules:`
>
> `- Under 125 characters.`
>
> `- Describe what is visible and relevant to the page, not everything in the frame.`
>
> `- Don't start with "image of" or "photo of".`
>
> `- Include visible text from the image if it matters.`
>
>
>
> `Give me one version. Then give a longer 2-sentence description I can use as a caption for screen reader users where more detail helps.`

### 5. Turn a photo into a listing description

> `I'm uploading photos of [item]. Write a listing description for [Etsy / eBay / Facebook Marketplace / my shop].`
>
>
>
> `Only describe what you can actually see in the photos: colour, material, visible wear, size cues. If something I'd normally state in a listing can't be seen (exact dimensions, brand, age, condition details), list it at the end under 'Details to confirm' instead of guessing.`
>
>
>
> `Tone: [plain and honest / warm and personal]. Length: [80-120] words.`

The "Details to confirm" list is what stops invented measurements. It's the same trick we use when we tell a model not to fabricate details in resume prompts.

### 6. Extract text or data from a photo

> `This photo shows [a whiteboard from a meeting / a receipt / a handwritten note / a slide]. Transcribe all legible text exactly as written. Mark anything you can't read with [unclear] rather than guessing.`
>
>
>
> `Then organise it as [a bullet list of action items / a table of date, vendor and amount / clean paragraphs].`

Check the output against the photo. Handwriting and low-light shots are where transcription slips.

## A worked example: vague versus specific

Take a phone photo of a handmade blue ceramic mug on a messy desk, with a laptop cable curling across the frame.

**Vague prompt:** "Make this photo look more professional."

What you typically get: a paragraph of general tips (use natural light, declutter, try a neutral background), or an edited image where the mug's glaze has shifted colour and the desk looks airbrushed. Nothing in the prompt said the glaze colour mattered, so nothing protected it.

**Specific prompt:**

> `I'm uploading a photo of a handmade ceramic mug for my Etsy shop listing. Keep the mug's shape, glaze colour (a deep blue with lighter speckles), and the handle exactly as they are.`
>
>
>
> `Change only this:`
>
> `- Replace the desk with a plain off-white surface.`
>
> `- Remove the laptop cable.`
>
> `- Soften the harsh shadow on the right.`
>
>
>
> `If you can't return an edited image, give me a numbered list of edits I can make in a photo editor.`

The answer now comes back scoped to three changes, with the glaze described as something to preserve. If your version can't edit images, you get a short edit list tied to this mug rather than generic advice. The difference isn't cleverness. It's that the second prompt left less to guess.

## Common mistakes that waste a turn

**Uploading a tiny or heavily compressed image.** Fine detail like small text or fabric texture gets lost. Send the original where you can.

**Asking for a style without a reference.** "Moody" and "cinematic" mean different things to different people. Describe it: "low-key lighting, deep shadows, warm highlights, muted greens." If you have a reference photo, upload both and say which is the target.

**Stacking ten changes in one go.** Edit in two or three rounds. After each, say what to keep from the last version. Models drift when the list gets long.

**Trusting identification or measurement.** Don't rely on a model to state exact dimensions, ages or authenticity from a photo. Treat those as claims you verify yourself.

**Skipping the follow-up.** If the result is close, don't rewrite the whole prompt. Reply with one correction: "Keep that, but the background is too grey; make it warmer white." Short corrections often work better than starting over. For more on tightening a draft prompt, see our guide to a [ChatGPT prompt optimizer](https://promptpal.app/blog/chatgpt-prompt-optimizer-improve-prompts).

## Making your photo prompts reusable

A good photo prompt is a template, not a one-off. The product cleanup prompt above works for the next fifty listings if you only swap the placeholders. Keeping it in a chat history means you'll never find it again.

Save the ones you use to a folder or a note, or put them in PromptPal's [prompt collections](https://promptpal.app/features/collections) so each task (listing photos, headshots, alt text) has its own set and anyone on your team can reuse it. If a template stops working after a model update, edit it once and everyone gets the fix.

If you also work with Gemini for image tasks, our [Gemini prompt templates](https://promptpal.app/blog/gemini-prompts-templates-work-images) cover the same ground with that model's quirks in mind. Wording that works in one model often needs a small tweak in another, so test before you assume.

## Adapting the templates to your own work

The six prompts above are starting points. Here's how to bend them.

- **Real estate or interiors:** In the "keep" line, list fixtures and room dimensions that must look accurate. In the "change" line, ask for brighter exposure and straightened verticals, not furniture swaps you couldn't deliver in person.
- **Food photography:** Keep the colour of the dish. Ask about plating, angle and light direction in the critique prompt rather than asking for the food itself to be altered.
- **Social media:** Add the platform and crop ratio to the purpose line, such as a 4:5 portrait crop for a feed post.
- **Team documentation:** For screenshots and whiteboard photos, the extraction prompt works best when you name the output format up front, so everyone's notes look the same.

Whatever the use, resist the urge to make the prompt clever. State the purpose, protect what matters, list the changes, name the output. That covers nearly everything.

## Next step

Browse the [public prompt library](https://promptpal.app/prompts/public) for ready-made templates by role and task, then save the ones you reuse so your best photo prompts are one click away next time.

---

Originally published at https://www.promptpal.app/blog/chatgpt-photo-prompts-templates
