Background Remover
Background Remover
Drop your image here or click to upload
JPG, PNG, WebP supported — up to 10 MB
Higher quality = larger model download (~40 MB / ~80 MB / ~160 MB). Cached after first use.
Upload an Image to Get Started
Drag and drop a JPG, PNG, or WebP image into the upload area, or click to browse your files. The AI will remove the background entirely in your browser — no data leaves your device.
AI-powered, 100% client-side — your images never leave your device
Removing the background from a photo used to require professional editing software, significant time, and a trained eye for masking and selection tools. Today, thanks to advances in machine learning and the ability to run neural network models directly in the browser, background removal is available to everyone — instantly, for free, and without uploading your images to any server. Our Background Remover uses the IS-Net deep learning model, the same technology powering professional-grade background removal services, running entirely on your device through WebAssembly. Unlike cloud-based tools that send your images to remote servers, every pixel of your photo is processed locally on your computer or smartphone. This means your personal photos, product images, ID pictures, and confidential documents never leave your device — a meaningful privacy advantage that most competitor tools cannot match. The tool supports JPG, PNG, and WebP input formats up to 10 MB, which covers the vast majority of everyday photography and product photography use cases. After background removal, you can download a transparent PNG, composite your subject onto a solid color of your choice, or place your subject in front of a completely different background image you provide. These background replacement options make the tool useful for e-commerce product photography, social media graphics, presentation slides, and creative photo projects. Three model quality options let you balance speed against precision. The Fast model (~40 MB) delivers quick results suitable for simple subjects like products on plain backgrounds. The Balanced model (~80 MB, default) provides excellent quality for most everyday use cases including portraits, animals, and complex subjects. The Best model (~160 MB) maximizes edge detail retention for subjects with fine hair, fur, or intricate outlines. After your first use, the model is cached by your browser so subsequent uses are near-instant regardless of which quality tier you select. Background removal has dozens of practical applications. E-commerce sellers use it to achieve the clean white or transparent backgrounds required by Amazon, eBay, Etsy, and other marketplaces. Photographers use it to composite portraits into different environments or remove distracting backgrounds from travel photos. Graphic designers use it to extract subjects for use in marketing materials, presentations, and social media posts. HR professionals use it to standardize headshots for company directories. Students use it to create clean visuals for school projects without needing expensive software. The tool also provides an Output Type option that exposes the underlying model's capabilities beyond simple foreground extraction. The Foreground mode (default) keeps your subject and makes the background transparent. The Mask mode generates a grayscale alpha matte showing exactly what the model detected as foreground versus background — useful for understanding model confidence, creating artistic effects, or further processing in other tools. For best results, use images where the subject is clearly separated from the background with good contrast. Well-lit photos on simple backgrounds give the most accurate results. The AI handles complex cases like hair, fur, and translucent elements remarkably well, though extremely fine details or subjects that blend into the background may require additional touch-up. Future versions of this tool will include a manual brush tool for precisely correcting any areas the AI handles imperfectly.
Understanding AI Background Removal
What Is AI Background Removal?
AI background removal is the process of automatically detecting and separating the main subject of a photograph from its background using a machine learning model. Traditional background removal required hand-drawing a selection mask around the subject — a time-consuming process requiring skill and patience. Modern AI models learn from millions of labeled images to identify subjects (people, animals, products, vehicles) and generate a pixel-perfect alpha matte that defines which areas should be transparent. The IS-Net model used in this tool is an Intermediate Supervision Network trained specifically for image matting tasks, achieving professional-quality results on a wide range of subjects and conditions. The model runs as an ONNX (Open Neural Network Exchange) file, a standardized format that can be executed efficiently in-browser via WebAssembly or WebGPU.
How Does It Work?
When you click Remove Background, the tool dynamically loads the selected ONNX model (downloaded from the IMG.LY CDN and cached by your browser for future visits). Your image is decoded and resized to the model's input resolution, then passed through the neural network which produces an alpha matte — a grayscale map where white pixels represent the foreground subject and black pixels represent the background. This matte is then applied to your original full-resolution image, making background pixels transparent. The result is encoded as a PNG file (the only common image format that supports full transparency) and displayed with a checkerboard pattern to indicate transparent areas. If you selected a background replacement option, the transparent result is composited onto your chosen color or background image using an HTML5 Canvas before download.
Why Does Client-Side Processing Matter?
When you upload an image to a cloud-based background removal service, your photo travels over the internet to a remote server, is processed, stored temporarily (sometimes for 24 hours or more), and then sent back to you. This raises privacy concerns, particularly for sensitive images: personal family photos, government-issued ID documents, medical images, proprietary product designs, or confidential business materials. Client-side processing eliminates this risk entirely because your image is never transmitted anywhere — the AI model comes to your data rather than your data going to the model. This is the same reason why password managers and encryption tools run locally rather than in the cloud. For professional use cases involving NDAs, GDPR compliance, or simply personal peace of mind, client-side processing is the objectively safer choice.
Limitations and Best Practices
AI background removal works best when the subject has clear visual contrast against the background and is well-lit with minimal motion blur. The model can struggle with certain challenging scenarios: subjects wearing clothing that closely matches the background color, transparent or semi-transparent objects like glass, subjects partially hidden behind other objects, images with very complex textures throughout, and extremely small subjects. Very large images (above 4000×4000 pixels) may slow processing significantly. The model is particularly accurate on people, animals, and common product categories because these appear most frequently in training data. Less common subjects or unusual perspectives may produce less accurate results. For difficult cases, try the Best quality model, as the larger network has more capacity to handle edge cases and fine details like individual strands of hair or wisps of fur.
How to Remove an Image Background
Upload Your Image
Drag and drop your JPG, PNG, or WebP image onto the upload zone, or click the zone to open a file picker. Images up to 10 MB are supported. A thumbnail preview will appear confirming your selection.
Choose Your Settings
Select a model quality (Balanced is recommended for most images), choose whether you want the foreground subject or just the alpha mask, and optionally select a background replacement — transparent, a solid color of your choice, or a custom background image.
Click Remove Background
Press the Remove Background button. On first use, the AI model will be downloaded (40–160 MB depending on quality) and cached by your browser for instant subsequent uses. A progress bar shows download and processing status.
Download or Copy Your Result
Once processing completes, your result appears on the right. Toggle between the original and result to verify quality. Click Download PNG to save the file, or Copy to place it directly on your clipboard for pasting into other applications.
Frequently Asked Questions
Does this tool upload my images to a server?
No — absolutely not. This background remover runs the AI model entirely inside your web browser using WebAssembly technology. Your image pixels are processed locally on your own device (computer, tablet, or smartphone) and never transmitted anywhere. The only network activity that occurs is the one-time download of the ONNX model file from a content delivery network (CDN), after which the model is cached by your browser. Your photos, including personal pictures, product images, ID documents, and any other sensitive material, remain completely private. This is a fundamental architectural difference from cloud-based tools like remove.bg or Photoroom, where your image is uploaded to their servers for processing.
Why does the first use take longer than subsequent uses?
On your first use, the AI model file must be downloaded to your device. Depending on which quality tier you select, this is approximately 40 MB (Fast), 80 MB (Balanced), or 160 MB (Best). This download is a one-time step — your browser caches the model file, so every subsequent use of the same quality tier loads the model instantly from your local cache. If you clear your browser cache or use the tool on a different device, the model will need to download again. On a typical broadband connection, the initial download takes 5–30 seconds. On your second visit, background removal typically completes in 2–10 seconds depending on your device's processing power.
What image types and sizes does the tool support?
The tool accepts JPG (JPEG), PNG, and WebP image files up to 10 MB in size. These formats cover the overwhelming majority of everyday photos and product images. HEIC and HEIF files (from iPhone cameras) are not currently supported — you can convert HEIC to JPG using your phone's share function or an online converter before using this tool. For best results, use images where the subject occupies a reasonable portion of the frame and has good contrast against the background. Very large images (beyond 4000×4000 pixels) may process slowly in-browser. The output is always a PNG file, which is the only widely supported format that preserves full transparency information.
How do I get the best background removal quality?
Several factors affect quality. First, use a well-lit image where the subject is clearly distinguishable from the background — high contrast between subject and background gives the AI the clearest signal. Second, choose the Best quality model for subjects with fine details like hair, fur, feathers, or translucent elements. Third, avoid images where the subject's clothing or fur closely matches the background color, as this is genuinely difficult even for the best AI models. Fourth, ensure the subject is in focus — blurry edges create ambiguity for the model. If the initial result has errors in small areas, try a different quality level. Manual touch-up brush tools (coming in a future update) will allow pixel-level corrections for the most demanding cases.
What is the Mask output type used for?
The Mask output type returns a grayscale PNG rather than a transparent image. In this mask, white pixels (value 255) represent areas the model identified as foreground subject, black pixels (value 0) represent background, and gray pixels (partial values) represent uncertain edge regions. This mask is useful for several advanced workflows: you can load the mask into Photoshop or GIMP as a layer mask for precise manual refinement, use it to create artistic effect such as halftone or threshold treatments, composite the subject in a way that your preferred software handles differently, or understand exactly how confident the model is at various edges of your subject. Most everyday users should use the default Foreground output type.
Can I replace the background with a custom color or photo?
Yes. The Background Replacement section offers three options. Transparent (default) gives you the subject on a transparent background, ideal for use in design software or on websites. Solid Color opens a color picker where you can choose any color — white is commonly required for e-commerce platforms like Amazon. Custom Image lets you upload any JPG, PNG, or WebP file to use as the new background. The tool composites your extracted subject in front of your chosen background using an HTML5 Canvas, and the result is what you download. For the most natural-looking composites, choose a background image with similar lighting direction and color temperature to your original photo.