Skip to content
Kordu Tools Kordu Tools

How to Remove Image Backgrounds for Free

Remove image backgrounds free in seconds using AI. No signup, no upload. Works for product shots, headshots, logos. Save as transparent PNG or WebP.

I
iyda
10 min read
background remover remove background transparent background ai image editing product photography

Cutting out a background used to mean hours with Photoshop’s Pen tool. Today, AI models trained on millions of images can do a cleaner job in under 10 seconds. E-commerce sellers, content creators, and designers all benefit from this shift. The bottleneck is no longer skill or software access. It’s knowing which tool to use and how to get clean results the first time.

Key Takeaways

  • AI background removal runs in your browser -- your images never leave your device.
  • Amazon and most major marketplaces require product images on a pure white background to be listed.
  • Always save with a transparent background as PNG or WebP -- JPEG cannot store transparency.
  • Clean, high-contrast source images with simple backgrounds consistently produce the best AI cutouts.
  • Hair, fur, glass, and transparent objects are the hardest cases for any AI background removal tool.

Remove Your Background Now

Drop your image below. The AI model runs entirely in your browser — nothing is uploaded to a server. First use downloads the model (around 5 MB), which is then cached for all future visits.

Try it Background Remover

Output Format

This browser will use the CPU path. First run may briefly feel heavy while the model warms up.

Click to upload or drag and drop

PNG, JPG, JPEG, WEBP up to 500MB

image compression

Why Does Background Removal Matter?

Background removal is one of the most common image editing tasks for online sellers, marketers, and content creators. Amazon explicitly requires product images on a pure white background for main listing photos (Amazon Seller Central Image Guidelines, 2025). The same requirement appears across Etsy, Shopify, and most major e-commerce platforms. A cluttered or distracting background shifts attention away from the product itself.

Beyond e-commerce, clean cutouts are useful in a wide range of everyday contexts.

  • Profile photos on LinkedIn and professional directories look sharper with a neutral or brand-coloured background replacing a busy office scene.
  • Presentations benefit from images that integrate with slide backgrounds rather than awkward white boxes.
  • Marketing materials like flyers, social cards, and banners need subjects isolated so designers can control the composition.
  • Logos and brand assets often need a transparent version for use on dark backgrounds.

product image formats

Why Manual Removal Is Slow

Photoshop’s Pen tool is still the gold standard for complex cutouts, but it takes years of practice to use well. Even an experienced designer spends 10-30 minutes on a single product photo with detailed edges. For a seller with 200 SKUs, that is weeks of work.

The Magic Wand and Quick Selection tools are faster but unreliable. They select based on colour similarity, which breaks down the moment the subject has similar tones to the background. Any photo shot indoors without studio lighting creates exactly that problem.

Outsourcing to a retoucher costs roughly $0.50-$5.00 per image depending on complexity (Retouching Pricing Guides, 2025). For a small business, that adds up quickly. AI tools have pushed the viable threshold down to cases that are genuinely difficult, like tangled hair or glass objects.

How AI Background Removal Works

Modern AI background removal uses a technique called semantic segmentation. The model assigns each pixel a label: foreground subject or background. It doesn’t rely on edge contrast alone. It has learned, from training on millions of images, what people, products, animals, and objects look like, and it applies that knowledge to your photo.

The model running in this tool belongs to the ISNet family of segmentation models. ISNet was designed specifically for high-resolution salient object detection, which is why it handles fine edge detail better than older approaches. It runs via ONNX Runtime in your browser, using WebAssembly or WebGPU depending on what your browser supports. Nothing is sent to a server.

Privacy by design

The AI model downloads once on first use and runs locally in your browser. Your images are never uploaded, stored, or processed on any server. The download is around 5 MB and is cached indefinitely in your browser.

Step-by-Step: Using the Background Remover

Step 1: Prepare your image

Start with the highest-quality source file available. The AI works at pixel level, so more resolution means more detail in the final cutout. A 2000px wide image is plenty for most uses. Clean up obvious issues before processing: crop to remove unnecessary empty space, correct the exposure if it’s very dark.

Step 2: Choose your output format

The tool offers two output options. PNG is lossless and universally supported. Use it for product photos, profile pictures, and anything you’ll edit further. WebP produces smaller files and also supports full transparency. Choose WebP if you’re optimising for web page load times.

PNG vs WebP for transparency

Step 3: Upload and process

Drag your image into the drop zone or click to browse. The model loads on first use, which takes a few seconds. Processing a typical photo runs in 5-15 seconds depending on image size and your device. A progress bar tracks both model loading and processing.

Step 4: Check the preview

Use the “Show Original” toggle to compare the result directly against the source image. Check the edges carefully, especially around hair, sleeves, and fine details. The checkerboard pattern in the preview represents transparency, so what you see is exactly what will download.

Step 5: Download

Click the download button to save. The file is named with -nobg appended to the original filename so you can tell it apart at a glance.

Getting the Best Results

Lighting and contrast

Flat, even lighting reduces shadows that bleed into the background. Shadows are ambiguous to the model: they belong to the subject physically but visually blend with the background. Softbox lighting or natural light from a window pointed at the subject (not behind it) minimises this.

Contrast between subject and background matters. A grey jacket photographed against a grey wall is harder than the same jacket on white. You don’t need a professional setup. A bedsheet or craft paper as a temporary backdrop works well for product shots.

Image sharpness and size

Blurry edges confuse the segmentation model. Shoot with adequate depth of field so the subject’s edges are sharp. Avoid heavy in-camera sharpening, which creates halos that look artificial in the cutout. File size affects processing time but not quality — the model downsamples internally if needed.

What to avoid

  • Transparent or translucent subjects: glass, water, netting, lace
  • Subjects that are the same colour as the background
  • Strong shadows cast across the subject-background boundary
  • Motion blur at the edges

Hair and fur take more work

Flyaway hair and fur are the hardest cases for any automated tool, including this one. The AI handles most hair edges well, but very fine, backlit strands on a light background may need a manual touch-up in a photo editor after downloading. For a quick fix, use an image editor’s “Refine Edge” brush on the downloaded PNG.

What to Do After Removing the Background

A transparent PNG is a starting point, not a finished asset. What you do next depends on how you’re using the image.

Add a solid colour background

For product listings that require a white background, open the downloaded PNG in any image editor (Photoshop, GIMP, Canva, Figma) and place a white layer beneath it. Flatten and export as JPEG. The result meets marketplace requirements and compresses well as JPEG since there’s no transparency to preserve.

Add a new background scene

Place the cutout over a background photo for composite images. The transparent PNG gives you clean edges to work with. Pay attention to lighting direction: the light in the background photo should match the light that fell on the subject in the original.

Compress before uploading

Transparent PNGs with large canvas sizes can be unexpectedly large because lossless compression doesn’t reduce the file as aggressively as lossy formats. After placing your subject on a white background and flattening, compress the resulting JPEG using our Image Compressor before uploading to your store or website.

compress images after editing

Why You Must Save as PNG (Not JPEG)

This is a common mistake that destroys the transparency you just created. JPEG was designed in 1992 for photographic images and does not include an alpha channel. There is no mechanism in the format to store which pixels are transparent. Saving a cutout as JPEG fills transparent areas with a solid colour, usually white or black, depending on the editor’s defaults.

PNG has supported full 8-bit alpha transparency since its original specification (PNG Specification, W3C, 1996). Every pixel stores a transparency value from 0 (fully transparent) to 255 (fully opaque), which is exactly what a background removal tool produces.

WebP also supports full alpha transparency and produces smaller files than PNG in most cases. It’s a practical choice when you’re done editing and just need to serve the image on a web page.

Format Transparency File Size Edit-friendly Best use
PNG Full alpha Larger Yes Editing, logos, product cutouts
WebP Full alpha Smaller Yes (limited tool support) Web display, page assets
JPEG None Smallest No (lossy) Flat photos, no transparency needed
GIF 1-bit (on/off only) Varies Not recommended Avoid -- use WebP instead

full format comparison

Use Case Tips by Context

E-commerce product photography

Most platforms require pure white backgrounds for primary images. After removing the background, add a white layer, flatten, and export as JPEG at quality 80-85. Compress the result to stay under 500 KB. For secondary images or lifestyle shots, you can keep transparency and layer over contextual backgrounds.

Profile photos and headshots

Most social platforms accept PNG or JPEG. If you’re uploading to LinkedIn or a personal website, add a solid brand colour or neutral grey background before exporting as JPEG. Keep the PNG version on hand for future use as a template.

Presentations and slide decks

PowerPoint and Google Slides both accept PNG with transparency. Drop the cutout directly onto a slide and it integrates with any slide background. Avoid JPEG for this use case — the white fill will be obvious against dark or coloured slides.

Logos and brand assets

Logos almost always need a transparent version. Deliver both: one on white background (JPEG for email/print) and one transparent (PNG for web, merchandise, dark backgrounds). Check the cutout at small sizes, where compression artefacts and rough edges show most.

Limitations: When AI Struggles

No tool handles every image equally well. Understanding the hard cases saves time and prevents frustration.

Transparent and semi-transparent objects. Glass, water, crystals, and sheer fabric are partially invisible — the model can’t separate what belongs to the subject from what belongs to the background underneath. Results are inconsistent. Manual editing with a photo editor is more reliable here.

Similar subject and background colours. If a person is wearing a grey coat in front of a grey wall, the model has very little information to distinguish them. Reshoot with better contrast if possible.

Complex scenes with multiple overlapping subjects. The tool is designed for single-subject isolation. Dense crowd scenes, full-table food photography, and wildlife shots in natural habitats are outside its reliable operating range.

Fur, feathers, and detailed hair. Fine strands blending into a light background remain the hardest challenge in segmentation. The model handles most hair well, but extreme backlit or fine wispy hair may show fringing or gaps.

Quick fix for edge fringing

If the cutout shows a thin colour fringe around the edges (usually the original background colour), many photo editors have a “Remove Color Matting” or “Defringe” option. In Photoshop: Layer menu - Matting - Defringe. Set the width to 1-2 pixels.

Frequently Asked Questions

What format should I save after removing the background?

Save as PNG to preserve full transparency. Every pixel stores its own opacity value, so the transparent areas remain invisible wherever you use the image. JPEG does not support transparency at all, so it fills transparent areas with a solid colour. WebP is a good alternative to PNG when file size matters and you’re done editing.

Can I remove the background from a photo with people?

Yes. Human subjects, including portraits and full-body shots, are among the most common and well-handled use cases. Single-person photos with reasonable contrast between the subject and background consistently produce clean results. Detailed hair is the most challenging part — fine flyaway strands may need a quick manual clean-up in a photo editor.

Why does the tool struggle with hair?

Hair is algorithmically difficult because individual strands are semi-transparent, extremely fine, and often blend with the background. The segmentation model works at pixel level, and a single pixel at a hair edge may contain both hair colour and background colour mixed together. The model has to make a judgment call on each ambiguous pixel. Shooting against a high-contrast background and ensuring the hair is sharply focused gives the model the most information to work with.

Why is the first run slower than subsequent ones?

The AI model file is around 5 MB and downloads on first use. Your browser caches it locally, so every run after the first skips the download entirely. Processing time after the model is loaded is typically 5-15 seconds depending on image dimensions and whether your device uses GPU or CPU inference.

Does this tool work for product images for Amazon?

Yes, with one additional step. Amazon’s primary image requirement is a pure white background, not just a transparent background. After removing the background, place the cutout on a white canvas in any image editor and export as JPEG. The AI handles the isolation. Adding the white background is a 30-second step in any editor.


For further reading on image formats and when to use each one, see the Image Formats Explained guide. To reduce file sizes after editing, the Image Compressor runs directly in your browser alongside this tool.

next logical step after removing background

Related articles