withoutbg 1.0.4

Python SDK for local and cloud background removal

withoutbg

Remove backgrounds in Python. Free locally, or use the cloud API.

Two modes that share the same API — run the open-weights model locally (free, private, offline) or call the cloud API (better quality, no GPU, pay per image). Switch with one line of code.

View Documentation →

Try it

pip install withoutbg

from withoutbg import WithoutBG

model = WithoutBG.open_weights()
model.remove_background("photo.jpg").save("result.png")

Choose your mode

	Local (open_weights())	Cloud (api())
Cost	Free forever	Pay per image
Quality	Good	Better (esp. hair, fur)
Privacy	Stays on your machine	Image sent to API
GPU required	No (CPU ONNX)	No
First-run setup	~495MB download, once	API key only
Best for	Offline, private, batch jobs	Products, occasional use

Compare results →

Need offline or private processing?   → Local
Processing a large batch?             → Local (pay the setup cost once, amortize across all images)
Building a product?                   → Cloud (better quality, zero infra overhead)
Occasional use, no setup tolerance?   → Cloud

Quick start

Local (withoutBG Open Weights Model):

from withoutbg import WithoutBG

model = WithoutBG.open_weights()
result = model.remove_background("input.jpg")  # returns PIL Image (RGBA)
result.save("output.png")

Cloud (withoutBG API):

from withoutbg import WithoutBG

# Pass api_key here, or set WITHOUTBG_API_KEY in the environment
model = WithoutBG.api(api_key="sk_your_key")
result = model.remove_background("input.jpg")
result.save("output.png")

Batch processing:

from withoutbg import WithoutBG

model = WithoutBG.open_weights()  # load once

images = ["photo1.jpg", "photo2.jpg", "photo3.jpg"]
results = model.remove_background_batch(images, output_dir="results/")

Progress callback:

def on_progress(value: float) -> None:
    print(f"{value * 100:.0f}%")

result = model.remove_background("photo.jpg", progress_callback=on_progress)

See examples/ for runnable scripts.

CLI

# Single image (local model)
withoutbg photo.jpg

# Batch
withoutbg ~/Photos/vacation/ --batch --output-dir ~/Photos/no-bg/

# Cloud API
export WITHOUTBG_API_KEY=sk_your_key
withoutbg photo.jpg --use-api

# JPEG output with white background fill
withoutbg portrait.jpg --format jpg --quality 95

withoutbg --help

Example outputs

See Local model results → See Cloud API results →

What gets returned

All methods return a PIL Image in RGBA mode:

result = model.remove_background("photo.jpg")  # PIL Image, RGBA

result.save("output.png")   # PNG — preserves transparency
result.save("output.webp")  # WebP — also supports transparency
result.save("output.jpg")   # JPEG — transparency is dropped silently

Configuration

Environment variable	Effect
WITHOUTBG_API_KEY	API key for Cloud mode (alternative to passing api_key=)
WITHOUTBG_MODEL_PATH	Path to a local .onnx file (skips Hugging Face download)

When using WITHOUTBG_MODEL_PATH, the sidecar metadata file (withoutbg-open-weights.onnx.json) must be in the same directory.

Performance

	Local	Cloud
First run	5–10s (~495MB download)	1–3s
Per image	2–5s	1–3s
RAM	~2GB	None
Disk	495MB (one-time cache)	None

Keep the model object alive across all images in a batch. Recreating it for every image reloads the weights each time.

Troubleshooting

Model download fails: The weights are pulled from Hugging Face on first run (~495MB). Check your connection, or set WITHOUTBG_MODEL_PATH to a local copy.

Out of memory: The local model uses ~2GB of RAM. Reduce batch size or switch to Cloud mode.

Import error:

which python
pip list | grep withoutbg
pip install withoutbg

API key rejected: Get a key at withoutbg.com. Set export WITHOUTBG_API_KEY=sk_your_key.

Migrating from older API names (WithoutBG.opensource(), ProAPI): see docs/MIGRATION.md.

Error handling

from withoutbg import WithoutBG, APIError, WithoutBGError

try:
    model = WithoutBG.api()
    result = model.remove_background("photo.jpg")
    result.save("output.png")
except APIError as e:
    print(f"API error: {e}")
except WithoutBGError as e:
    print(f"Processing error: {e}")

Model

The withoutBG Open Weights Model is a unified WBGNet ONNX graph hosted at withoutbg/withoutbg-openweights-onnx. It runs depth estimation, segmentation, matting, and refinement in a single inference pass at up to 768px output resolution. Licensed under the withoutBG Open Model License (Apache 2.0 for withoutBG portions; Meta DINOv3 License for DINOv3 backbone weights). Built with DINOv3.

Development

uv sync --extra dev
# or: pip install -e ".[dev]"

make test-fast    # fast unit tests
make quality      # lint + format + type check
make test         # full suite (downloads model on first run)

See CONTRIBUTING.md for the full guide.

Related projects

Need a browser UI or HTTP API instead of Python?

withoutbg-inference — Docker images (CPU + GPU), FastAPI inference service, and optional web UI built on the same open-weights model.

docker run --rm -p 8080:8080 withoutbg/withoutbg-openweights-v3-app-cpu

License

This Python SDK is licensed under Apache License 2.0. See LICENSE.

The withoutBG Open Weights Model is a composite artifact with additional terms for embedded DINOv3 weights. See the withoutBG Open Model License, LICENSE-DINOv3, and NOTICE.

Third-party components

• DINOv3 (Meta): Meta DINOv3 License — backbone weights in the Open Weights Model
• Depth Anything V2: Apache 2.0
• ISNet: Apache 2.0

See THIRD_PARTY_LICENSES.md for complete attribution.

Support

• Bugs / questions: GitHub Issues
• Discussion: GitHub Discussions
• Commercial: [e-mail]
• Security: [e-mail] — see SECURITY.md