Why BiRefNet
Older background removers were built on general segmentation models that output a rough, low-resolution mask. BiRefNet was designed specifically for separating foreground from background at high resolution, and it looks at global context and fine local detail at the same time. In practice that means thin structures — hair, bicycle spokes, plant stems — survive.
What happens after the AI
A model's mask is only half the job. We resize it to your image with an edge-aware guided filter, so the edge follows real detail in your photo. Then we re-estimate the colour of each semi-transparent pixel, removing the tint of the old background. That last step is why a person shot against a green wall doesn't end up with a green outline.
Last updated November 1, 2026