Skip to main content

Image Tools · No account required

Remove Unwanted Objects from Images with a Brush

Brush over unwanted objects, fill them from nearby image patches, or touch up with a clone brush. Free, local photo cleanup without AI.

Instructions, examples and limitations ↓

What this tool does

Paint a selection over a small unwanted object, then replace it using intact patches from the same image. A separate clone brush lets you choose the source yourself. Everything runs in your browser without an AI model or photo-upload service.

Enable JavaScript to use the local processing controls. Instructions and examples are available below.

Practical guide

Remove Unwanted Objects from Images with a Brush

By MultiZest · Updated

How to use this tool

  1. Choose a JPG, PNG, WebP, AVIF or BMP image up to 40 MB, 4 megapixels and 4,096 pixels per side, or load the example.
  2. Use the Remove brush to cover the complete object and its shadow. Zoom in, adjust brush size and use Restore selection to correct the marked area.
  3. Choose Texture repair for repeated detail, Smooth background for a plain surface, or Mirror repair when the opposite side of a symmetrical subject supplies the missing structure. For texture repair, optionally mark clean replacement material with Source brush.
  4. Choose Remove selection and wait for the local fill. Compare with View original; Undo and Redo let you review an edit. Use the clone brush for a remaining difficult edge.
  5. Fill or clear any remaining selection, prepare a PNG or JPG download, and inspect the saved copy at full size.

Worked example

Load the sky example and cover the orange circle with a brush slightly wider than its edge. Remove selection replaces the marked circle with sky patches. The example demonstrates a small object on a simple background; a face or building hidden by a large object is a harder task.

Use a source that supplies the missing surface

Source brush marks acceptable replacement material in green. Texture repair then searches only clean patches wholly inside those marks, excluding the selected object. For a distraction over a shirt, choose intact matching fabric rather than a nearby face. For an object across a wall and clothing, repair those surfaces separately instead of asking one source to supply both.

For a front-facing portrait, Mirror repair can copy an unobstructed neck or collar from the opposite side. Place the blue mirror line through the subject’s center and keep the selected object on the obstructed side. The editor refuses a mirrored source that overlaps the selection. This is a copy of visible material; it is unsuitable when pose, clothing, lighting or writing differs between sides. Smooth background instead extends surrounding colors and does not recreate fabric or anatomy.

A repair brush needs a useful surrounding area

Look at what should remain behind the object. Sky, a plain wall, grass or another repeated texture gives the filler useful material. A painted selection over a unique face, sign or complicated architectural feature does not provide enough information to reconstruct the hidden scene reliably.

Cover the object completely, including its faint shadow and antialiased outline. Leaving a sliver outside the selection preserves that sliver in the output. Avoid brushing over more background than necessary; intact nearby pixels help the algorithm choose a plausible replacement.

Remove a small distraction before a large one

For a photo of a product on a plain table, begin with a small crumb or isolated mark. Inspect the first result before selecting several regions. Separate passes make a poor repair easier to identify and undo.

Automatic filling accepts up to 150,000 marked pixels and 25% of the image per pass. The full image keeps its original pixel dimensions, subject to the editor’s input limits. Larger source images should be resized deliberately in the crop and resize tool before opening this workspace; they are not silently reduced.

Choose the clone source around the edge you want to keep

Click Choose clone source, then select an intact area with the same texture and lighting as the destination. A green marker identifies it. Switch to the clone brush and paint over the distraction. The source follows your movement within each stroke, beginning from the selected source point again for the next stroke.

Use a softer edge for a gradual wall or sky repair and a harder edge for a clearly defined line. Choose another source as the texture changes. A patch of brick from the wrong row or light direction can look convincing in isolation and misaligned in the complete picture.

Compare the actual repair at two sizes

Use View original to see what changed; editing pauses while that view is active. Then return to the edited view and look closely for repeated patterns, seams and leftover edges. Zooming changes the workspace display, not the exported image dimensions.

Also inspect the photo at its final display size. A minor seam invisible in a profile thumbnail may be obvious in a large product image. Undo restores the image and selection from an earlier action, and Redo reapplies that state. Reset image returns to the original opened in this session.

What no-AI filling does under the brush

Texture repair refines matching patches at several image sizes and combines compatible overlapping copies instead of accepting the first small patch at each point. Match patch lighting adjusts local colors and blends the repair boundary. A very smooth surrounding area can use color interpolation without an expensive texture search. These calculations do not understand or invent a hidden face, hand or garment.

The calculation runs in a browser worker so the editor can show progress and offer Stop filling. Stopping retains the last completed image and the selection. Mirror repair copies the clean opposite side, and the clone brush copies from a source you select. None loads trained weights or calls an image-processing API.

Save an editable source and a distribution copy

PNG saves the current pixels losslessly and retains transparency outside the repaired area. JPG uses 95% quality and a white background behind transparent pixels. Purple selection marks, green sources, the blue mirror line and the cursor are excluded from both formats.

A browser preview is not a saved file. Prepare the download, save it, then open that copy in the receiving app. The workspace has no recoverable account history; refreshing can discard edits. Undo history is limited to eight actions and a 48 MB snapshot budget, so large images retain fewer actions.

Choose a repair that fits the purpose of the photo

For a casual listing photo, removing a small background distraction can help the reader see the item. For evidence, identity documents or records, preserve the unedited original and follow the receiving organization’s rules. A tidy repair changes what the photograph shows.

A visible cover and a plausible replacement solve different problems. Use image censorship when a private detail must be obscured. Use object cleanup when the goal is an ordinary visual repair, and review the entire image before publication.

Use the saved image to judge the final result

Check the downloaded picture in its final layout and at full pixel size. Look for repeated texture, an abrupt color change and small parts of the original object. A clean selection is helpful, but it does not establish that the generated repair is visually convincing.

Keep the original privately and name the edited file by purpose. If the first fill is unhelpful, undo it and try a more precise selection or a deliberate clone source. Repeatedly applying the same selection is not a reliable way to restore hidden detail.

Questions and troubleshooting

Does this object remover use AI?

No. It uses ordinary patch matching and pixel copying inside your browser. The clone brush uses the source you choose. There is no AI model download or image-upload processing service.

Can it remove any object perfectly?

Results depend on the visible surroundings. Small objects on simple or repeated backgrounds work best. Hidden faces, text and large unique structures cannot be reliably recovered from missing pixels. Use clone touch-ups, another source photo or a dedicated editor where needed.

Is this the same as removing a background?

Object cleanup replaces selected content with image material. Background removal creates a cutout with transparency. The separate portrait background remover uses an AI model and serves that different task.

Why did part of the object remain?

Its outline or shadow may have been outside the selection. Undo, zoom in and cover the complete unwanted region. Automatic filling changes selected pixels only.

Can I work without a mouse?

Touch and pen input are supported. With keyboard focus on the canvas, arrow keys move the cursor, Shift increases movement to ten pixels, and Enter or Space applies a dab or selects a clone source. Toolbar buttons provide Undo and Redo.

Can I fill a transparent area?

Automatic filling expects an opaque selected area and intact opaque donors. Use the clone brush for transparent edges; PNG preserves the final alpha channel.

Why should I keep the original after a good repair?

A flattened export has no editable repair layer or recoverable object. The original lets you adjust the selection, choose another source or meet a later requirement without reconstructing discarded pixels.

Continue with a related tool

More detailed guides

Sources and further reading

Found an error or an unclear instruction? Report it. Read our editorial policy and privacy policy.