Object Remover

Paint over a person, object or text in a photo and an AI fills the gap — free, in your browser, nothing uploaded.

100% private — runs entirely in your browser. Your data is processed on your device and never sent to the internet.

Click to upload or drop a photo here

JPG, PNG or WebP — used in your browser, never uploaded

The AI model downloads once on the first removal: about 28 MB (Fast) or 208 MB (Best), then it is kept in your browser. Works best on small objects, people and text on simple backgrounds. Slow on phones. Not for removing other people's watermarks or copyright marks.

About the Object Remover

This free object remover takes a person, an object or a piece of text out of a photo. You paint over it with a brush, and an AI model fills the gap with a matching background. You get a clean photo to download as PNG or JPG.

Everything runs in your browser. Your photo is never uploaded to a server. The AI model downloads once, and your browser keeps it for next time.

Features

  • Brush and rub-out — paint over the thing to remove, and rub out any mistake before you start
  • Brush size and zoom — a size slider and zoom let you paint small objects with care
  • Two AI models — Fast (28 MB) for quick jobs, Best quality (208 MB) for smoother fills
  • Sharp photo — only the painted area changes, so the rest of the photo keeps every pixel
  • Undo and repeat — undo a removal, or paint again to remove more from the new result
  • Download PNG or JPG — one click, no watermark, no sign up

Why it is free and private

Most online erasers send your photo to a company's server. This one does not. The AI model runs on your own device, so a private photo stays private. There is no account, no watermark and no limit.

Powered by MI-GAN (MIT licence, Picsart AI Research) and LaMa (Apache-2.0, ONNX port by Carve), run with onnxruntime-web (MIT).

The first removal downloads the model. After that, your browser keeps it.

What it works best on

The model fills a gap by looking at what is around it. So it does well when the background is simple or has a pattern.

  • Small objects — a bin, a cable, a spot, a sign, a stray bottle on a lawn
  • People in the distance — a photobomber on a beach, sky, wall or street
  • Text and dates — a date stamp, a caption or a label on a plain background

It does not do well on big objects in front of complex scenes, or on faces and hands. The fill can look blurry or odd there. Paint a part at a time, and use Undo when the result is not good.

Use it fairly

Use this tool for your own photos, or photos you have the right to edit. It is not for removing other people's watermarks or copyright marks.

Tips for a good result

  • Paint a little outside the object, so its shadow and edges are covered
  • Remove a big object in several small steps, not one big stroke
  • Use the Best quality model for large smooth areas like sky, sand or walls
  • On a phone, choose Fast and use small photos. The Best model needs a lot of memory

Frequently asked questions

Is the object remover free?

Yes. There is no sign up, no watermark and no limit on photos. You can download the result as PNG or JPG.

Is my photo uploaded?

No. The photo and the result stay on your device. The only thing that downloads is the AI model, once, from a public model host.

How big is the first download?

The Fast model is about 28 MB. The Best quality model is about 208 MB. The page shows both sizes before you press the button. After the first time, your browser keeps the model.

How large can my photo be?

Photos up to 4096 px on the long side are used as they are. Bigger photos are scaled down to 4096 px when you open them. The AI only works on a small window around your paint, so the rest of the photo stays sharp.

Does it work on a phone?

Yes, but it is slow, and the Best model may run out of memory. Use the Fast model on a phone. You can paint with your finger, and the Move button scrolls a zoomed photo.

Will it work without internet?

Only after the model has downloaded once. Your browser keeps it, so later removals can work with a poor connection. The page itself still needs to load.

Can I use the results for business?

Yes, the pictures are yours. The AI models are open source: MI-GAN is under the MIT licence and LaMa is under Apache-2.0, and both allow free use.

Related tools

How to use it

  1. Open your photo — click the upload area or drop a JPG, PNG or WebP file. No photo at hand? Press Try a sample photo and paint over the bin.
  2. Choose the AI model — Fast is best for most photos. Choose Best quality for large smooth areas.
  3. Paint over the object — use the brush and set the size. Cover the whole object and a little of its edge and shadow.
  4. Press Remove — the first time, the model downloads (28 MB or 208 MB). Then the gap is filled in a few seconds.
  5. Check and repeat — hold the eye button to compare. Use Undo, or paint again to remove more.
  6. Download — pick PNG or JPG and press Download.

Video Tutorial

We add new tools regularly — subscribe on YouTube to be notified when each one goes live.

More tools

View all

Recommended Apps

View all