Remove Background

Batch remove backgrounds from images using rembg (Python library).

When This Skill Activates

•User says "remove background", "make transparent"
•User has images that need transparent backgrounds
•User wants to batch process multiple images

Prerequisites

rembg must be installed:

bash

pip install rembg
# or
pipx install rembg

Verify installation:

bash

rembg --version

How To Use

Single Image

bash

rembg i INPUT.png OUTPUT.png

Batch Process Directory

IMPORTANT: Never output to the same file - use a separate output directory to avoid corruption.

bash

# Create backup first
mkdir -p backup && cp *.png backup/

# Process to separate folder
mkdir -p processed
for img in *.png; do
  echo "Processing: $img"
  rembg i "$img" "processed/$img"
done

# Verify results, then replace originals
mv processed/*.png . && rmdir processed

Process Entire Folder (rembg native)

bash

rembg p INPUT_DIR OUTPUT_DIR

Output Format

•Input: Any image (PNG, JPG, WEBP)
•Output: PNG with RGBA (alpha channel for transparency)

Example Workflows

Web Assets

bash

cd ~/projects/my-app/public/images
mkdir -p transparent
for img in *.png; do rembg i "$img" "transparent/$img"; done

Product Photos

bash

cd ~/products/photos
mkdir -p cutout
rembg p . cutout/

Models Available

rembg supports different AI models:

Model	Quality	Speed	Use Case
u2net (default)	Good	Fast	General purpose
u2net_human_seg	Better	Medium	People/portraits
silueta	Good	Fast	Similar to u2net
isnet-general-use	Best	Slower	Complex backgrounds, pixel art, illustrations

Recommended for pixel art/illustrations: isnet-general-use

Specify model:

bash

rembg i -m isnet-general-use input.png output.png

Troubleshooting

Issue	Solution
Empty output file	Don't use same input/output path - use separate folder
Slow first run	Model downloading, wait for it
Memory error	Process fewer images at once
Poor quality	Try different model with `-m` flag

Pro Tips

•Always backup before batch processing
•Process to separate folder first, verify, then replace
•Use -a flag to keep alpha if image already has transparency
•Post-process in image editor for fine edges if needed