How to remove people from a photo

What erases cleanly and what does not

Before you mark anything, it helps to know what the software actually does. An object remover does not reveal what was behind the person in your photo: that information was never captured. It looks at the pixels around the hole and paints something plausible to fill it. When the surroundings are simple and repetitive, plausible and correct are the same thing. When they are unique, the fill is an educated guess.

What disappears without a trace:

• A small figure against open sky, sea, sand, snow or a lawn. The texture repeats, so the fill has plenty of examples to copy.
• Someone walking along a plain rendered wall, a stone pavement or a gravel path with a regular pattern.
• Power lines, phone cables and thin wires crossing a sky. They are narrow, so the model only has to invent a strip a few pixels wide, and the sky on either side tells it exactly what belongs there.
• Litter on a beach, a parked bike against a hedge, a road sign on a field edge, a bin in a corner.

What leaves an invention:

• A large group of tourists standing directly in front of a cathedral door, a fresco or a shopfront with lettering. The fill will produce a door, a wall or letters, but not the ones that were there.
• A person covering a third of the frame. The bigger the hole, the less real context sits around it, and the more the result looks soft or repetitive.
• A car hiding the only part of a building you cared about, such as a doorway or a plaque.
• Someone overlapping the person you want to keep: hands on shoulders, hair in front of a face. The edge between the two is where the fill struggles most.

A quick rule: look at the area you want to erase and ask whether you could recreate it by copying a nearby patch of the same photo. If yes, the result will usually be clean. If the answer is no, because what is hidden is one of a kind, expect a believable but invented background. For a holiday photo that may be fine; for a picture of a specific monument, it may not be. In that case it is often better to crop, or to accept one or two small figures and remove only the ones that distract.

How to paint the mask to remove objects from an image

The mask is the red area you paint over what you want gone. Almost every patchy result comes from the mask, not from the model, so it is worth a minute of care.

Cover the object and its traces. A person standing in the sun casts a shadow on the ground. A boat on calm water has a reflection. A car leaves a dark patch beneath it. If you paint only the object, the shadow or reflection stays behind with nothing attached to it, and the eye spots it immediately. Paint the shadow and the reflection as part of the same shape.

Leave a generous margin. Paint a little outside the outline, roughly the width of a finger on screen at normal zoom. The edges of a person in a photo are never perfectly sharp: there is motion blur, a faint halo from compression and a few pixels of colour from their clothes bleeding into the background. If the mask stops exactly at the outline, those pixels remain and the fill builds around a ghost.

But do not eat into what you want to keep. A margin is good on the side facing background; it is bad on the side touching a column, a railing or another person. Where the object meets something you want to preserve, follow that edge closely and keep the margin on the other sides.

Work by zones rather than painting everything at once. If you want to remove people from a photo of a busy square, it is tempting to paint all fifteen of them in one go. That gives the model fifteen holes and less clean context around each one, and where two holes are close together it has to invent a large area with almost nothing real to borrow from. Painting one group, processing, and then painting the next group on the result means each fill can draw on the already-cleaned surroundings. The results are steadier and, if one area goes wrong, you only need to redo that area.

A practical order: start with the largest or most central object, then the medium ones, and finish with the small background figures and stray details such as a sign or a bin.

Common mistakes when you remove tourists from a photo (and wires)

Most failed attempts to remove tourists from a photo come down to a handful of mistakes. They are easy to avoid once you know where to look.

1. Leaving the shadow. The person is gone, but a grey silhouette still lies across the pavement. Long afternoon shadows can stretch several times the height of the person, so follow them all the way to the end.
2. Leaving the foot. Feet, ankles and the bottom of a trouser leg are often slightly darker than the ground and blend into it at screen size. Zoom in on the base of every figure you erase. A single shoe left on a cobbled street is the classic giveaway.
3. Leaving a bag, a selfie stick or a pushchair. Objects carried by a person are easy to miss because your eye reads them as part of the background. Check the area around the hands.
4. Painting across a straight line. If the mask covers part of a railing, a skirting board, a kerb, a window frame or the horizon, the fill has to reconnect the line on both sides. Sometimes it does; sometimes it bends it, breaks it or shifts it by a few pixels. The fix is to keep the mask off the line wherever you can, or to split the work: erase the figure on one side of the line first, then the other.
5. Masks that are too tight. A faint outline of the removed person remains, especially around hair and against a bright sky.
6. Masks that are too big. Painting half the frame to remove one person forces the model to invent far more than necessary, and the result looks smeared.
7. Trying to remove a person who overlaps the subject. If a stranger's arm is behind your friend's head, the fill has to invent part of your friend's hair. Expect to need a second, smaller pass, or decide that the arm can stay.

Cables deserve a separate note. When you want to remove wires from a photo, paint each wire as a thin, continuous stroke slightly wider than the cable itself, including the spot where it joins a pole or a wall. Breaking the stroke into short segments leaves small dots of cable between them.

One big pass or several small ones, and checking at 100 %

Several small passes work better when:

• There are many objects scattered across the frame, such as tourists in a square or boats in a harbour.
• Two objects are close together and their masks would merge into one large hole.
• The object sits next to a straight line or a detail you want to keep, and you want to treat each side separately.
• The first attempt left a small trace, such as a shadow fragment or the tip of a shoe. A second, small mask on that spot is quicker than redoing everything.

One pass is enough when the object is single, isolated and surrounded by simple texture: a lone walker on a beach, a sign in a field, a wire across a clear sky.

Check at 100 % before you download. At screen size, almost any fill looks convincing, because the photo is being shown at a fraction of its real resolution. Problems hide in the detail. Zoom to 100 %, where one image pixel equals one screen pixel, and look specifically at:

• Repeated patches: the same paving stone, cloud or leaf appearing twice side by side.
• Texture that is too smooth compared with the area around it, which shows as a soft blob in an otherwise sharp photo.
• Lines that do not meet: railings, roof edges, horizons and kerbs.
• Grain and noise. If your photo was taken at night or at high ISO, the filled area may look cleaner than the rest. That mismatch is subtle on screen and obvious in print.

If the grain mismatch bothers you, it can help to remove noise from the photo first and erase afterwards, so the whole image has the same clean texture. And if the photo is small and you plan to print it, erase first and upscale the image as the last step, so the enlargement works on the finished picture rather than on the patches.

Step by step: remove objects from a photo in Airpic

Here is how to remove objects from an image in Airpic, from upload to download. There is no account to create and nothing to install; the mask editor runs in your browser.

1. Open the tool. Go to the page to remove objects from a photo. You can drag the file in, pick it from your device or paste it with Ctrl+V. JPG, PNG, WebP, TIFF, HEIC and AVIF are accepted, so a photo straight from an iPhone works as it is. The EXIF orientation is applied and the colour profile is kept.
2. Paint the mask. The photo appears in the editor. Choose a brush size slightly wider than the object, then paint over it, including its shadow, its reflection and anything it is carrying. Keep a small margin towards the background and stay tight against edges you want to keep.
3. Start with one zone. If there are several people or objects, paint the biggest or most central group first rather than the whole frame.
4. Process. Processing and viewing the result is free. The before and after appear in a comparator, so you can drag the divider across the area you erased.
5. Inspect at 100 %. Zoom in on the filled area and check for leftover shadows, feet, broken lines and repeated textures, as described above.
6. Next zone. If anything else needs to go, paint the next area and process again, so each fill builds on the cleaned result.
7. Download. When you are happy, download the full-resolution file. One full download a month is free; after that you use credits, which do not expire. The tool page has the current details.

Your files are deleted one hour after download, and after one day at most, and they are never used to train models. Processing runs on GPUs in the European Union. If you want to understand why we work without accounts or subscriptions, read why Airpic works this way.

Two more ideas once the photo is clean: if the picture is going on a website or a listing, compress the image so it loads quickly; it is free. If it is an old print with dust and scratches rather than a holiday shot, the guide to restoring old photos with AI covers a better route. For a crowded travel photo, though, a few careful passes in the object remover are usually all it takes.

FAQ

Keep reading