In short
Use this tool for one unwanted person, sign, small object, or distraction at a time. Mark the item with the brush, wait for the repair, and download only after you have checked the result. A narrow, deliberate selection usually gives the editor more useful surrounding pixels than a large one.
How to remove an object from a photo
Object removal is most reliable when you treat it like a small repair, not a crop. Start by choosing the original JPEG or PNG rather than a compressed copy from a chat app. The original preserves more texture in walls, pavement, foliage, fabric, and skies—the evidence the editor uses to fill the selected region. Once the photo appears above, paint directly over the person or object you want gone. The brush is deliberately manual: it lets you decide exactly what is unwanted instead of asking an automatic detector to guess the subject of your image.
Cover the complete item, including a thin edge that touches the background. If a bag strap, chair leg, reflection, or cast shadow is clearly caused by the object, include that detail too. Do not paint a large halo around it just to be safe. The unselected background is the strongest instruction you can give the repair: it tells the model whether it should extend grass, tile, wallpaper, water, or the side of another object. A tight outline makes that instruction clearer and helps preserve nearby faces, lettering, and architecture.
Select Remove selection when the red painted area covers the distraction. The result appears on this page so you can compare it with your original before downloading. Zoom in mentally on the border where the object used to meet the background. Look for repeated textures that jump, a line that changes direction, or a shadow that no longer agrees with the light in the scene. If the repair feels unnatural, clear the selection and make a second, more precise pass. A smaller mark is often better than repeatedly processing the same broad patch.
Where photo cleanup works best
Open backgrounds
Sky, sea, sand, grass, pavement, painted walls, and uncluttered interiors give the tool room to infer what belongs behind an item. These are good first images to try.
One clear distraction
A passerby at the edge of a travel photo, a bin beside a building, or a small object on a table is easier than several overlapping subjects. Work through separate objects one at a time.
Visible context
Results improve when the same material is visible beside the marked area. If an object hides all of a unique detail, such as a face or a logo, there is less real context to reconstruct.
Good results still need a human review. The study behind this page found that photo-tool users notice when an export does not match what they saw in the editor, when a repair changes identity-sensitive details, or when a supposedly automatic tool takes away control. That is why this page keeps the original visible until you choose a file, uses a painted selection rather than a hidden automatic decision, and makes the final download an explicit step. The repair is a suggestion for your image—not a fact about what was behind the object.
The honest limits
This tool cannot recover pixels that were never captured. When someone blocks a face, a hand, a product label, or the only part of a building, the output is a generated reconstruction rather than the original hidden detail. Fine text, repeated railings, mirrors, transparent objects, dense crowds, and strong geometric patterns can expose seams or invented details. Very low-resolution images also give the repair less texture to work with.
- Do not use a generated result as evidence, authentication, or a record of what was originally present.
- Do not remove copyright, ownership, licensing, safety, or attribution marks from media you do not have the right to edit.
- Keep the original photo. If the image matters for a sale, report, identity, health, legal issue, or memory, compare the result carefully before relying on it.
Tips for a more natural result
Match the brush to the item rather than the entire surrounding scene. For a person standing on a path, include the shoes and the small shadow beneath them but leave the rest of the path unpainted. For an object against a wall, follow the object’s outline and preserve any wall texture around it. If the image contains a repeating pattern—tiles, bookshelves, fence posts, bricks, or windows—make the selection as narrow as possible, because a wide selection forces the tool to invent a longer run of that pattern.
Use separate passes when distractions are far apart. Removing a stranger from the background of a family photo is a different task from cleaning a sign near the foreground. A separate pass lets you inspect each repaired area and leaves more reference material in place for the next one. It is also safer for people and meaningful objects near the selection: the less unrelated content you paint over, the less the generated region has reason to touch it.
Finally, check the downloaded file, not only the on-page preview. Open it at the size you intend to share and inspect the repair where it meets the original pixels. This final check is worth doing because export is part of the product experience: an image that appears acceptable in a small preview can show softness, artifacts, or a mismatched texture when viewed in a gallery, printed, or sent to another person.
Frequently asked questions
Can I remove a person from a photo?
Yes. Upload the photo, paint over the person you want to remove, and let the tool rebuild the selected area. It works best when there is visible background around the person for the editor to use as context.
Can I remove an object from a picture?
Yes. Brush over the unwanted object rather than the whole image. Include its edge and any shadow or reflection that belongs to it, then check the result at full size before downloading.
Why does a complex background need more than one try?
The tool has to infer pixels that were hidden by the object. Repeating patterns, faces, text, railings, and tight architecture give it less unambiguous context, so a smaller or more carefully placed selection can produce a better result.
Will removing an object change the whole photo?
The edit is focused on the region you select, but image generation can affect nearby pixels while it blends the repair. Keep the selection tight and compare the downloaded image with your original when exact detail matters.
Does this work for watermarks, text, and shadows?
It can help with small unwanted text or objects in ordinary scenes. Do not use it to remove ownership notices, licensing information, or marks from media you do not have permission to edit. For shadows, include only the shadow caused by the removed object.
Which photo types work best?
JPEG and PNG photos work best when the selected item has clear surrounding context: an empty patch of wall, sky, pavement, grass, water, or a repeating texture. Start with the original image instead of a tiny screenshot whenever possible.