Pure Blush | Background Removal 1.0

Art Institute of Pure Blush Background Removal

Free local model · ISNet segmentation in your browser · no queue · popup API included

Drop a photo here, click to browse, or press Ctrl+V from clipboard

PNG · JPG · WebP · only your own works or content with consent

Models load on first run and stay cached in your browser.

Compare

Model

Background

FINE-TUNE — RECOMMENDED DEFAULTS

Mask threshold 50%

Edge feather 2px

Expand mask 0px

Download format

2 models — size · specialization

Fast…
44MB · ISNet int8
DEFAULT
Quality…
88MB · ISNet fp16
EDGES

API for your website

API idle. Another website can open this remover as a popup and send a photo for processing.

For AI agents & bots

Machine-readable spec of the popup API. Requires a JavaScript-capable client — the neural engine runs locally in the browser, there is no REST endpoint.

Download check & diagnostics

Every model download is byte-checked. If anything fails, copy the log below and send it to us.

No. 01 Is my photo uploaded anywhere?
No. The model downloads once into your browser cache, then everything runs locally on your device (WebGPU when available, CPU fallback otherwise). Close the tab and nothing remains on any server — because nothing was ever sent.
No. 02 Fast or Quality — which model?
Fast (44MB) is the recommended default: tiny download, quick on phones, clean masks for people, products and pets. Quality (88MB) keeps finer hair and semi-transparent edges — pick it for portraits and tricky fur.
No. 03 What do the fine-tune sliders do?
Threshold sets where the mask cuts: higher keeps only confident foreground (no halo), lower keeps wispy hair. Feather softens the edge by a few pixels. Expand grows (+) or shrinks (−) the mask to swallow leftover background fringe or reclaim edge detail. Sliders re-render instantly from the cached mask — the model runs only once.
No. 04 Model licenses — what must I know?
Both models are MIT-licensed ISNet general-use weights (imgly base, community int8 quantization) — free for personal and commercial use. Full list with links is in the license panel below.
Fast · ISNet int8 · 44MBMIT. source · base weights: imgly/isnet-general-onnx (MIT)
Quality · ISNet fp16 · 88MBMIT. source
Free AI background remover · transparent PNG maker · remove background online · no signup · WebGPU background removal

How AI background removal works

The background remover segments your photo with the ISNet neural network at 1024px resolution and exports a transparent PNG — right in your browser. Two quality levels cover quick cutouts and detailed edges with hair and fur, while threshold, expand and feather sliders re-render the mask instantly without re-running the model.

Private by design

Both models run locally via WebGPU with CPU fallback. Your photos are never uploaded, stored, or used for training. Please edit only your own works or content you have consent to use.

Engine: ISNet 1x segmentation at 1024px, running locally via onnxruntime-web (WebGPU with CPU fallback). Mask post-processing (threshold, expand, feather) re-renders instantly without re-running the model.
Art Institute of Pure Blush — Hardstrasse 201, 8005 Zurich · Prime Tower · pureblush.dev