| name | tools:background-removal |
| description | Remove backgrounds from images locally with Python and rembg, producing transparent PNGs. Works fully offline (no API, no login) on a single image file or a whole folder of images. Use when someone says remove background, remove bg, background removal, rembg, cut out image, make transparent, transparent background, transparent png, cutout, isolate subject, product photo cutout, or wants a PNG with the background stripped. Triggers on either a single file or a batch folder. |
Strip image backgrounds locally with Python + rembg, writing transparent PNGs. Runs offline (no API, no login) on a single file or a whole folder.
Prerequisites
uv — verify with uv --version. If missing, install it: https://docs.astral.sh/uv/getting-started/installation/ (or winget install astral-sh.uv on Windows). uv handles Python 3.10+ and all dependencies automatically via the script's inline metadata.- First run downloads the model (
u2net, ~170 MB) into the rembg cache. Every run after that is offline.
Usage
The bundled script lives at ${CLAUDE_PLUGIN_ROOT}/skills/background-removal/scripts/remove_bg.py. Invoke it with uv run — dependencies (rembg, pillow) resolve and cache automatically:
Single image:
uv run "${CLAUDE_PLUGIN_ROOT}/skills/background-removal/scripts/remove_bg.py" path/to/image.jpg
Folder (batch all images in it):
uv run "${CLAUDE_PLUGIN_ROOT}/skills/background-removal/scripts/remove_bg.py" path/to/folder
Multiple inputs / options:
uv run "${CLAUDE_PLUGIN_ROOT}/skills/background-removal/scripts/remove_bg.py" a.jpg b.png folder/ -o out_dir -m isnet-general-use
Flags:
-o, --output-dir DIR — write all PNGs into DIR instead of beside each source.-m, --model NAME — rembg model (u2net default; try isnet-general-use for higher quality, u2netp for speed).
On Windows/PowerShell, quote paths that contain spaces. ${CLAUDE_PLUGIN_ROOT} resolves to the installed plugin root.
Behavior
- Output is a transparent
.png written beside each source by default (foo.jpg → foo.png).
- A
.png input is written as <name>_nobg.png so the original is never overwritten.
- Supported inputs:
.jpg .jpeg .png .webp .bmp .tif .tiff. Non-images are skipped with a notice.
- Each image reports
[ok] or [fail]; the run ends with a Done: N/M images. summary. One bad file does not abort the batch.
Self-healing
If the script misbehaves (wrong output, an unhandled error, a model that won't load), fix scripts/remove_bg.py and this SKILL.md directly rather than only patching the current run.
