Skip to content

AI Photo Editor: Upload a Photo, Describe the Change

Chat shortcut: /editphoto

With the AI photo editor, you upload a photo, describe the change in one sentence, and the image model makes it, with no layers, masks or prior experience.

With Account

Loading…

Your photo as the starting point

Unlike an image generator, this tool starts from your existing photo: it is handed to the image model together with your instruction, and the model computes a new image from it. The aim is to change only the part you describe: a different background, a removed person, a blue shirt instead of a red one, evening light instead of midday sun or a watercolor look. The remaining areas are meant to stay, but may differ slightly.

You write the instruction as you would to a retoucher: verb first, then the target area and the result, for example "Replace the background with a white studio wall" or "Remove the car on the left, everything else unchanged". If something must stay, such as the face or a label, say so explicitly. If the result is not quite right, send a short follow-up instruction in the same chat instead of starting over.

People use it for product photos that need a clean background, for professional profile photos (on LinkedIn, for example), for real estate photos with a gray sky, for social media posts in a particular style or for keepsake photos where a distracting detail should disappear.

How it works

  1. 1

    Upload an image

    Upload a JPG, PNG or WebP into the chat with your account; ideally a sharp original, not a screenshot.

  2. 2

    Describe the change

    Write in one sentence what should change and what must stay the same, for example "Replace the sky with blue sky, house and garden unchanged".

  3. 3

    Check the result

    Look at the image at full size, especially hands, hair, text, edges and the areas that should have stayed unchanged.

  4. 4

    Refine or save

    Download the image or send a short follow-up instruction until it fits.

Typical uses

Product photos

Set the background to white or a fitting scene, remove distracting props.

Professional profile photo

A calm background, soft light and a button-down shirt instead of a T-shirt, for LinkedIn and the like, no photo studio needed.

Real estate and vacation photos

Replace a gray sky with a blue one, remove a trash can or passersby.

Social media

Turn a photo into a comic, watercolor or vintage style for a post.

Keepsake photos

Remove distracting details or add color where it's missing.

Good to know

  • You can upload JPG, PNG or WebP; for iPhone photos in HEIC format, use Convert HEIC to JPG first. Image upload requires an account.
  • Each edit costs a fixed amount of tokens, shown before you send.
  • One change per instruction usually gives the better results. Make several changes one after another in the same chat.
  • The model computes a new image from the photo and the instruction; areas that should stay may differ slightly. Image models struggle with hands, long text, logos, reflections and fine hair edges; with heavy changes a face can change. Check the result before you use it.
  • You can write the instruction in plain English; photography terms such as "backlight", "telephoto lens" or "studio flash" help the model.
  • You need the rights to the original image; other people, logos and third-party brands remain protected. The photo is transmitted to the provider of the image model (Google or OpenAI) for editing; what the provider stores is governed by its terms. Edited images stay in your history until you delete the conversation.
  • For standard tasks there are separate tools with fixed settings: Background remover, Remove Object from Photo, Expand image and Text Remover.

Frequently asked questions

Which changes are possible?

For example a new background, removed or added objects, different clothing or colors, a different time of day or a different style. Describe the change as precisely as possible and say what should stay.

How do I write a good instruction?

Start with a verb, name the area and the desired result, and say what should stay the same: "Change the shirt to dark blue, face and background unchanged". One change per instruction usually works best.

What if the result isn't right?

Send a short follow-up instruction in the same chat that names only the problem, such as "softer shadow" or "hair as in the original". Each run costs tokens again.

Is the face preserved?

With small changes such as background, clothing or light, usually yes. Still write "face unchanged" in the instruction and check the result, because the model recomputes the whole image and with heavy changes features can shift.

Can the AI change text in the image?

Short words such as a price or a name often work; long sentences are error-prone. Put the new text in quotation marks and check the result.

Which formats can I upload?

JPG, PNG and WebP, with an account. Convert HEIC photos from an iPhone first with Convert HEIC to JPG.

What does an edit cost?

Each edited image costs a fixed amount of tokens, shown before you send. You need an account for the image upload; the credit comes from your daily allowance, packages or subscription.

Loading…