Remove image background

Modern background removal uses a salient-object segmentation model (rembg / U²-Net family) to predict a per-pixel mask, then applies it as alpha.

What works well

  • Single subject with clear silhouette (person, product on plain surface).
  • Reasonable contrast against the background.

Where it struggles

  • Hair wisps, glass, motion blur, transparent objects.
  • Multiple overlapping subjects.

Try it: Remove background. PNG output with real transparency.

← Back to Learn