Remove the background from your photos with a segmentation model that runs in your own browser — no servers, no limits, no watermarks
This tool uses AI semantic segmentation to automatically distinguish an image's main subject (a foreground person or object) from its background, and generates a cutout with transparency — without needing to manually select edges like you would in a traditional photo editor.
The model (a segmentation classifier from Google's MediaPipe family, ~250KB) runs entirely in your browser via WebAssembly and WebGL. It analyzes the image pixel by pixel and calculates, for each one, the probability it belongs to the main subject or the background — that probability map is used as a transparency (alpha) channel for the cutout. No image is uploaded to a server: all the analysis happens on your device.
🔒 Why this matters for your privacy: most free "AI background remover" tools online process your photo on a cloud server — meaning the image (sometimes of a document, product, or person) travels over the internet and stays, at least temporarily, in a third party's hands. This tool never sends your image anywhere.
🚀 How it works, step by step
Upload your image
Drag or select a JPG, PNG or WEBP file. If it's very large, we automatically resize it first so the browser can process it without running out of memory.
AI analyzes the image
The first time, the browser downloads the model (~250KB). It then identifies which pixels belong to the main subject and which belong to the background.
Choose the output background
Transparent (PNG), white, or black. You can change this option as many times as you want without reprocessing the image — the switch is instant.
Download the result
Get your background-free image directly on your device. It never went through any server.
💡 Real use cases
- E-commerce product photos: isolate the product on a white or transparent background for your catalog or marketplace listing.
- Profile pictures and avatars: cut out your photo to use over any background, on resumes, social media or business cards.
- Presentations and design: extract elements from a photo to compose them into a design, presentation or collage.
- Thumbnails: separate the subject from the background to create cleaner, more consistent thumbnails.
⚠️ Limitations (honest transparency)
- Optimized for people and objects with clear edges: the model ("selfie segmenter") was mainly trained to recognize foreground people. With very small, transparent objects or very fuzzy edges (for example, loose hair blowing in the wind, smoke, glass), the cutout may not be perfect.
- One main subject at a time: if the image has several separate people or objects, the model treats them all as "foreground" together — it doesn't let you pick one in particular.
- It's not manual editing: if you need to retouch specific edges with pixel-level precision, this tool gives you a good starting point, but doesn't replace a manual cutout tool for very demanding cases.
- Input size limited to 1600px per side for now, so processing works reliably in any browser, including mobile.
❓ Frequently asked questions
Do I need internet to use this tool?
Only the first time, to download the AI model (~250KB). Once downloaded, your browser caches it and you can keep using the tool offline as long as you don't clear the site's data.
Is it really free, with no usage limit?
Yes. Since all processing happens on your device (not on a server of ours), there's no variable cost per use that we'd need to charge you for. There's no image limit and no watermarks.
Are my photos uploaded to any server?
No. The image never leaves your browser: it's loaded, processed and downloaded locally. You can verify this yourself by opening your browser's developer tools (Network tab) and checking that the image isn't sent anywhere during processing.
Does it work well with product photos, not just people?
The model is mainly trained on people ("selfie segmenter"), but in practice it works reasonably well with any subject that has clear contrast against the background. For products with very irregular or transparent shapes, results may vary.
Can I change the background color after processing the image?
Yes. Switching between transparent, white or black is instant and doesn't require re-running the AI model — it just recomposes the already-processed image with the new background.
Why do hair or very detailed areas sometimes have imperfect edges?
The model generates a per-pixel confidence mask, not a perfect selection; areas with very fine detail (loose hair, lace, smoke) are inherently hard to segment precisely, even for larger and heavier AI models than this one.