Toolkit  /  Remove background

Remove background

AI-powered background removal that runs entirely in your browser. Coming soon — get notified when it launches.

Files never leave your browser
Coming soon
This tool is in development
We're building this carefully so it works as well as you'd want. Drop your email and we'll let you know the moment it's ready.

Background removal has historically been a tedious manual process — pick out the subject in Photoshop, trace the edges, refine the mask. Modern AI models (like U²-Net and BiRefNet) can do this automatically with high accuracy. But every existing online tool requires uploading your images to a server.

We're building a background remover that does the AI inference directly in your browser. Your product photos, your portraits, your design work — none of it leaves your device. The model runs locally using WebGPU or WebGL acceleration. It's coming soon.

Why privacy matters for background removal

Think about what people typically remove backgrounds from:

  • Product photos for e-commerce listings — work-in-progress designs you don't want competitors to see
  • Headshots and portraits with personal context
  • UI mockups and screenshots containing internal company information
  • Identity documents that need to be cropped to a portrait
  • Medical or sensitive images in healthcare workflows

For all of these, the standard "upload to a server" approach is at minimum uncomfortable and often unacceptable. Many enterprises explicitly prohibit uploading to third-party services. Doing the AI inference locally solves the problem entirely.

How it will work

When you drop an image, the browser will:

  1. Download the AI model file (a one-time ~80 MB download, cached after that) directly to your device.
  2. Run the model on your image using WebGPU (modern browsers) or WebGL (fallback), producing a precise mask of the foreground subject.
  3. Composite the result with a transparent background and output as PNG or WebP.

This means no upload, but also a small initial wait for the model download. Subsequent images process in 1–3 seconds depending on resolution and your device.

Release timeline

We're testing the model right now. The challenge isn't the AI — open-source models like RMBG-1.4 and BiRefNet are excellent — but making them fast enough to be usable in a browser. We're targeting under 5 seconds per image on a typical laptop.

Drop your email below to be notified when it launches. In the meantime, you can use any of our other tools — all live, all browser-based.

Common questions

Things people actually ask.

We're aiming for Q2 2026. Sign up above to be notified.

Yes. Same as all our other tools — free, no watermarks, no signup, unlimited use.

We're targeting accuracy matching the major paid services (remove.bg, Canva, Adobe). Modern open-source models like BiRefNet match or exceed commercial quality.

Yes, but slower than on a laptop. Modern phones (2022+) handle this fine. Older phones may take 10+ seconds per image.

Yes — drop dozens at once and they'll process in sequence. Each image is independent.