You’ve got a photo that’s nearly right. There’s a stranger wandering into the back of your holiday snap, or the light’s gone flat and grey, or you wish the shot was a bit wider so you could actually see the whole building. This is exactly the kind of thing AI photo editing tools are good at now, and you don’t need Photoshop skills to do it.

What you’ll be able to do

There are four edits most people actually want, and current tools handle all of them reasonably well.

Removing people or objects, so the bin bag, the photobomber, or the power line disappears and the space fills in convincingly. Fixing light, brightening a dark face or evening out a harsh shadow without making the whole thing look fake. Expanding the frame, adding believable extra scenery around the edges of a photo that was cropped too tight. And combining two photos, taking a person from one shot and placing them into another.

None of this is perfect. The tools guess what should fill a gap, and sometimes the guess is obviously wrong; a bit of stone path that doesn’t line up, a shadow pointing the wrong way. But for casual use, holiday photos, family pictures, things you want to share rather than publish professionally, it works, and it’s quick.

Before you start

Gather your photos first. Work from the original file if you can, not a screenshot of it or a version already squashed by WhatsApp. Quality in, quality out.

Think about which job you’re doing, because it changes which tool suits you. Removing a small object from a simple background is easy for almost anything. Removing a person standing in front of complicated detail, like a patterned wall or a crowd, is harder and needs a better tool. Expanding a frame works best when there’s a similar area nearby for the AI to copy the style from, so a beach or sky is easier than expanding into a room full of furniture.

Keep the original file. Always. Save your edit as a copy, not an overwrite, so if the AI makes a mess of it you haven’t lost the source.

And a word on ethics, because it matters here more than in most guides on this site. If you’re removing a person from a photo, that’s usually fine, it’s your photo. But editing a photo of someone else, changing what they’re wearing, where they are, what they’re doing, and then presenting it as real, that can cause real harm, and in some contexts it’s against the law. If you alter a photo in a way that changes what it shows, say so when you share it. Don’t let an edited image pass as an unaltered one.

Get set up

Where you start depends on what you’re doing and what you’ve already got.

If you’re on an Android phone with Google Photos, Magic Editor and Magic Eraser are built in and handle removal and reframing directly in the app. Availability and exact features vary by device and plan, so check what you’ve got before relying on it.

If you’ve got a recent iPhone, Apple Photos has a Clean Up tool that removes distracting objects and people from the background. It’s quick, on-device, and needs no extra sign-up.

If you want more control, or you’re combining two separate photos, ChatGPT or Gemini are worth trying. You upload a photo, describe what you want changed in plain English, and it edits the image and gives you a new version. This is the most flexible option for combining shots because you can describe exactly where each element should go. Claude and Copilot are strong for a lot of tasks but image editing from an uploaded photo isn’t their main strength; they’re better used for text, planning, or describing an image than for producing an edited one.

If you already use Adobe Photoshop, Generative Fill, Generative Expand and the Remove tool do all four jobs well, with the most precise control of any option here. It’s the heavier tool though, and there’s a learning curve.

Canva is a good middle ground: Magic Eraser, Magic Grab and its background remover are simple, browser-based, and enough for most casual edits without installing anything.

For background removal specifically, remove.bg does one job and does it fast: upload a photo, get back a clean cutout with the background gone. Handy if you’re prepping an image to combine into another scene.

None of these are wrong choices. Pick based on what device you’re already holding.

Try it yourself

Start small. One edit at a time, check it, then move on. Trying to remove a person, fix the light, and expand the frame all in one go usually gives worse results than doing each separately.

To remove something, upload the photo to ChatGPT or Gemini and describe exactly what should go and what should stay:

Here is a photo. Remove the person standing on the far right of the image, and fill in the background naturally so it looks like they were never there. Don't change anything else in the photo.

To fix the light, be specific about what’s wrong rather than just saying “make it better”:

This photo is too dark on the left side of the face. Brighten just that area so the lighting looks even and natural, without changing the colours elsewhere in the photo.

To combine two photos, upload both in the same message and describe the placement clearly:

I've uploaded two photos. Take the person from the first photo and place them standing next to the tree on the right-hand side of the second photo. Match the lighting and shadows so it looks like one photo, not two combined.

Small, clear instructions work better than vague ones. Say where, say what stays the same, say what should change. If the tool asks a follow-up question, answer it rather than starting again from scratch.

Check the result

Don’t just glance at the thumbnail and move on. Zoom in.

For a removal, check the area where the object used to be. Look for smudging, repeated patterns (a fence or brick wall that suddenly doesn’t line up), or a shadow that’s still there even though the object isn’t. Look also at the edges, sometimes a bit of the removed thing gets left behind, a stray hand or shoe.

For lighting fixes, check that skin tones still look like skin, not orange or grey. Compare the fixed area to an untouched part of the photo in similar light; if they don’t match, ask the tool to try again with a smaller, more specific area.

For an expanded frame, look at where the original photo ends and the new bit begins. Straight lines are the giveaway, horizons, walls, roads. If they kink or blur at the join, it hasn’t worked.

For combined photos, check the shadows point the same direction on both people or objects, check the scale looks right (nobody looks like a giant next to a house), and check the edges around the combined person for a faint outline or colour mismatch. This is the one that goes wrong most often, so look hardest here.

If it looks right at a glance but something feels off, trust that. Zoom to 100% and look properly.

If it doesn’t work

First attempt not right? Try being more specific rather than starting over completely. Instead of “remove the person,” try “remove the person in the red jacket standing near the lamppost, and match the pavement texture behind them.” More detail usually gets a better second attempt.

If a tool keeps failing on a hard background, try a different tool: something that struggles with a busy crowd scene in ChatGPT might do better in Photoshop’s Generative Fill, which was built specifically for this.

On what not to hand over: don’t upload photos of other people if you don’t have their permission to edit or share the result, even if the edit seems harmless. Don’t upload identity documents, screenshots containing personal information, or anything with visible addresses, number plates, or financial details, some tools may retain uploaded content depending on their settings and policies, so treat anything you upload as potentially not private. Check the privacy policy of whichever tool you’re using if you’re unsure, and avoid uploading images you wouldn’t be comfortable being stored somewhere.

And don’t pass off a heavily edited photo as an untouched one, particularly if it involves another real person. That’s the line between a fun edit and something that can genuinely hurt someone.

Keep exploring

If you want to build an image from scratch rather than edit an existing one, see “How to get AI to create an image”. If you’re turning a photo into something designed, like a flyer or event announcement, “How to get AI to make a poster” covers that. And if you’ve got a photo and want the AI to describe what’s in it rather than change it, that’s “How to get AI to tell me what is in a photo”.

Sources and review notes

Review date: 15 September 2026 — every source above was opened and checked against the text on that date.