Most background removers make you upload your photo to someone else’s server and hand back a watermarked, downscaled result unless you pay. This one works the other way round: the AI model (BRIA’s RMBG-1.4, in a quantized ONNX build) downloads to your browser once and runs locally with WebAssembly. Your photo never leaves the device — there is no server for it to go to — and the result is a full-resolution transparent PNG with no watermark.
Using it takes three steps: drop a photo onto the page (or pick a file, or paste straight from the clipboard), wait a few seconds while the model separates subject from background, and download the PNG. An original-versus-result comparison helps you check edges before you save, and two built-in sample photos let you see the quality before you trust it with your own picture.
The trade-off is the first visit: the model is about 44 MB, so it downloads with a progress bar and is then stored in your browser’s cache — later visits start almost instantly. Processing typically finishes in seconds on a computer and in about 10–20 seconds on a phone.
How to remove a background
- Drop a JPG, PNG or WebP photo onto the tool — or click to browse, or paste an image straight from the clipboard.
- Wait for the one-time model download (about 44 MB, first visit only); the tool shows its progress and remembers the model for next time.
- Let the AI run: it separates the subject from the background on your device, usually in seconds (10–20 seconds on phones — keep the tab open).
- Compare the cut-out against the original and check the edges, especially around hair and loose clothing.
- Download the transparent PNG at full resolution — no watermark, nothing to unlock.
How it works, honestly
On-device AI has real advantages and one real cost. Both are listed here so you know exactly what happens when you use the tool.
| Aspect | What happens | Good to know |
|---|---|---|
| The model | BRIA RMBG-1.4 (quantized) runs via ONNX Runtime Web and WebAssembly | A purpose-built background model — cleaner edges than general-purpose segmenters |
| Your photo | Decoded, processed and exported entirely on your device | Never uploaded; closing the tab leaves nothing behind |
| The one download | Model weights (~44 MB) from Hugging Face’s CDN, runtime from jsDelivr, with mirror fallbacks | Cached by the browser afterwards, so repeat visits start fast |
| Input | JPG, PNG or WebP — drop, file picker or clipboard paste | Two sample photos are built in if you just want to try it |
| Output | Transparent PNG at the photo’s full resolution | No watermark, no downscaling, ready for design tools and marketplaces |
| Speed | Seconds on a typical computer; about 10–20 seconds on a phone | The first run also includes the model download |
The interface is available in English, Simplified Chinese, Traditional Chinese, Japanese and Korean, and can be switched inside the tool.
Frequently asked questions
Is my photo uploaded to a server?
No — and that is the point of running the model on-device. Your photo is decoded and processed in your browser tab and the PNG is created locally; nothing about your image is sent anywhere. The only downloads are the AI model and its runtime library, fetched from public CDNs on the first visit.
Why does the first use take longer?
The AI model is about 44 MB and has to download once, with a progress bar so you can see it moving. Your browser then caches it (the tool shows “ready — cached”), so on later visits the remover starts almost instantly and the photo itself usually takes only seconds.
What do I get at the end?
A transparent PNG at your photo’s full resolution — no watermark and no downscaling. Drop it straight into a design, a marketplace listing, a presentation or on top of a new background colour in any image editor.
Which photos work best?
Photos with a clearly defined subject and some contrast against the background give the cleanest cut-outs — portraits, products and pets are ideal. Very fine detail (loose hair against a busy background, transparent objects like glass) is the hardest for any AI remover; zoom in and check the edge comparison before you rely on the result.
Is it really free? Can I use the results commercially?
The tool is free to use with no sign-up and no watermark. One honest caveat: the RMBG-1.4 model is provided by BRIA free for non-commercial use, so if you need cut-outs for a commercial project, check BRIA’s licence terms for the model before relying on it.
Which browsers does it work in?
Any modern browser with WebAssembly — current Chrome, Edge, Firefox and Safari on desktop and mobile. The tool tells you if WebAssembly is unavailable. On phones, keep the tab in the foreground while the AI runs; switching away can pause the computation.
Related tools
This tool in other languages: English · 简体中文 · 繁體中文 · 日本語 · 한국어 · हिन्दी · Español · Português · Deutsch · Français · Русский