DEV Community

KunStudio
KunStudio

Posted on • Originally published at bg-remover-ks.pages.dev

I built a background remover that keeps the fine hair edges

#webdev #ai #showdev #javascript

Background removal is a solved problem right up until the image has hair, fur, or fine detail, and then most tools smear it into a halo. I built BG Remover to get a clean, HD transparent PNG with crisp edges, including the hard parts like flyaway hair, and to do it per image without a subscription.

It is aimed at the practical cases: product photos, portraits, and logos that need a transparent cutout to drop onto any background.

Why I built it

The two common options are a free tool that outputs a low-res result with rough edges, or a subscription service you pay monthly whether you cut out one image or a hundred. For someone who occasionally needs a clean cutout for a product listing or a portrait, both are a bad deal. I wanted HD quality and per-image pricing.

How it works technically

  • Hosting: Cloudflare Pages with Functions. Static front end, serverless image processing, fast globally.
  • Model: It uses fal.ai's BiRefNet for the segmentation. BiRefNet is a strong matting model specifically because it handles fine boundaries, so hair and fur come out clean instead of as a blurry outline.
  • Output: A transparent HD PNG, ready to place on any background.
  • Payment: PayPal, priced per image. The single-image tier is $2.99. No subscription and no monthly lock-in.
  • Price validation: The per-image charge is confirmed server-side before the processed PNG is returned, so the price can't be changed in the browser.

Using it in three steps

  1. Upload your image. A product, a person, or a logo.
  2. It cuts it out. Clean HD edges, including fine hair and detail, in seconds.
  3. Download the PNG. A transparent PNG, ready to drop onto any background.

Try it here: bg-remover-ks.pages.dev

What I'd tell another builder

The differentiator in a "solved" category is the edge case, literally. Anyone can remove a background from a photo shot against a white wall. The reason to pick one tool over another is what happens at the hairline. Choosing a matting model built for fine boundaries, rather than a generic segmentation model, is what makes the output usable instead of just fast.

Top comments (0)