Paint over anything you want gone — a photobomber, a trash can, power lines, text — and an AI fills the gap with believable background. Runs entirely in your browser.
Drop a photo here
or
JPG, PNG, WebP · the AI model (~27 MB) downloads once, then everything runs on your device
No photo handy? Try a sample imageThe AI Object Remover erases things you wish were not in a photo: a stranger in the background, power lines across a sky, a trash can at the edge of the frame, a watermark of your own you want gone, a spill on the carpet. Paint over the object with the brush, click Remove selected, and an AI inpainting model reconstructs the background that was hidden behind it. The result looks like the object was never there, not like it was smudged out.
Everything happens on your device. The inpainting model (MI-GAN, an ICCV 2023 research model from Picsart AI Research) downloads to your browser once, about 27 MB, and then every edit runs locally through WebAssembly. Your photo is never uploaded, never queued on a server, and never logged. That matters for exactly the photos people most want to clean up: family shots, home listings, IDs on a desk, screenshots with private details.
Everything works without an account and without uploading your photo to a server you do not control.
Drop your photo into the tool, paint over the object with the brush, and click Remove. An AI inpainting model reconstructs the background behind it, so the object disappears instead of leaving a hole. There is no watermark, no signup, and no upload.
Yes. Brush over the person, including their shadow, and the AI fills the space with a continuation of the scene behind them. For people who take up a large part of the frame, remove them in a few smaller passes rather than one giant stroke for the cleanest result.
Here, yes, because your photo never goes online. The AI model downloads to your browser once (about 27 MB) and every edit runs on your own device. Nothing you load or erase is uploaded, queued, or logged on a server.
The technique is called inpainting. A neural network trained on millions of photos looks at the pixels around your selection and predicts what the covered area most plausibly looked like, then paints that prediction in. It works best when the background is texture, sky, grass, walls, or repeating patterns.
Inpainting predicts the background, and very large or very detailed areas are harder to predict. Two fixes usually solve it: erase big objects in several smaller passes instead of one huge selection, and keep the brush tight to the object so the AI keeps more real background to learn from.