AI portrait cutout · Browser processing

AI Background Remover

Remove a portrait background in your browser with an open-source AI matting model. Preview transparency or a solid background and download a PNG. Results vary by image and are best for people.

Remove an image background

Run an AI portrait-matting model in your browser, then preview transparency or add a new background.

AI model · Local inference
1

Upload a portrait

This free model is optimized for people and portrait-style subjects. It may struggle with products, pets, complex scenes or fine transparent objects. The model files download on first use; processing runs in your browser.

2

Run background removal

The first run loads the open-source model and can take a little while depending on your connection and device.

Choose a portrait image to begin.

3

Preview and customize

Processed previewNot processed yet
Your cutout will appear here after processing.
4

Download your result

PNG preserves transparency when the transparent checkerboard option is selected. Choosing a solid preview background exports a flattened image.

Free background remover for portrait images

This browser-based background remover uses the open-source MODNet portrait-matting model to estimate a foreground mask. You can preview the cutout over transparency or a solid color, then export a PNG. It is especially intended for portraits and people; it is not a guarantee of perfect results for every image type.

How to remove a background

  1. Upload a JPG, PNG or WebP image with a clearly visible person or portrait subject.
  2. Choose “Remove background” and allow the model files to load on first use.
  3. Wait while the browser runs the model locally.
  4. Inspect edges, hair and small details in the preview.
  5. Keep transparency or choose a solid background, then download the PNG.

Privacy, performance and quality

The image is processed in the browser; the first use requires downloading the model and runtime files from their hosting source. Model performance depends on device capability, subject placement, contrast and image complexity. The underlying Xenova/modnet model card lists the model under Apache-2.0 and describes it as portrait matting. Read the model card. For commercial deployment, keep the relevant license notices and model attribution available, and recheck model and runtime terms before launch.

For product photos, pets, busy backgrounds or fine transparent objects, review the output carefully. This tool is a convenient free option, not a promise of studio-grade segmentation on every image.

Frequently asked questions

Does it work offline?

After model files are downloaded and cached it may work without another download, but the first run requires a connection. Browser cache behavior varies.

Why is the first run slow?

The browser must download the model and initialize the machine-learning runtime. Performance also depends on your CPU, GPU and available memory.

Will it remove any background?

The selected model is optimized for portrait matting. Results can be weaker for products, pets, groups, transparent objects and complex scenes.

Does PNG preserve transparency?

Yes, when you select the transparent checkerboard option. A solid background is flattened into the downloaded PNG.

Compress the result ↗