Browser-onlyNo uploadsAlways free
About Us·Contact

100+ tools · No signup · Always free

AI-Powered Portrait Tool

Background Blur Tool

Blur photo backgrounds instantly with AI-powered portrait effects — fully browser-based and privacy-friendly.

This tool blurs the background of a portrait while keeping the subject sharp, producing the shallow depth-of-field look of a portrait lens. Subject segmentation runs on an AI model inside your browser, and blur strength is adjustable with a before-and-after comparison slider. Nothing is uploaded.

AI Segmentation4 Blur StylesBefore/After CompareFull-Res ExportNo Upload

Portrait vs Standard Blur — What's the Difference?

Compare blur modes to choose the right effect for your photo

FeatureStandardPortraitBokehSoft
Blur amountMediumMedium + darkStrongSubtle
Background colourNaturalSlightly darkNaturalBright
Edge sharpnessCleanCleanCleanClean
Best forAny photoLinkedIn/proCreative/artSubtle edit
DSLR lookModerateStrongStrongLow

Best Portrait Blur Tips

Get professional-looking results every time

Centre the Subject
AI segmentation is most accurate when the person is clearly separated from the background and positioned in the frame without heavy cropping.
Good Lighting Helps
Well-lit portraits with clear contrast between subject and background produce cleaner segmentation edges and a more professional-looking result.
Tune Edge Softness
Start with Edge Softness at 4–6. Too low gives an artificial cut-out look; too high creates a halo effect around the person.
Portrait Mode for Realism
The Portrait blur style desaturates and darkens the background slightly, mimicking the look of a real camera with a wide aperture — more natural than plain Gaussian blur.
Square Photos for Social
Export at 1:1 for Instagram posts. Crop before blurring for best results — the AI performs better on portraits cropped above the waist.
Use PNG for Precision
Export as PNG when you need to preserve fine edge detail, especially for hair and transparent elements. JPG is fine for general social sharing.

About This Tool

Technology, privacy, and how it works

MediaPipe Selfie Segmentation

This tool uses Google's MediaPipe Selfie Segmentation — the same AI model used in Google Meet, Zoom, and other tools for virtual backgrounds. It accurately separates people from backgrounds in real time, running entirely in your browser via WebGL.

100% Private

No image data is ever sent to a server. The AI model (~3 MB) is downloaded once from a CDN and then cached locally. All segmentation and blur compositing happens on your device using HTML Canvas and WebGL.

How the Blur Works

The AI generates a segmentation mask (a greyscale image where white = person, black = background). The tool then composites the original subject over a separately blurred version of the image, using the mask as a stencil. Edge softness feathers this boundary for a natural look.

Full-Resolution Export

The preview canvas is scaled for performance, but the download always renders at your original image's full pixel dimensions. Blur radius is scaled proportionally so the visual effect looks identical at any resolution.

How to use

1

Upload a portrait

Add a photo with a clear subject. The AI segmentation model loads into the page on first use, which takes a few seconds.

2

Wait for subject detection

The model separates the subject from the background and generates a mask. Everything runs in your browser — the photo is not uploaded.

3

Adjust the blur

Use the slider to set blur strength. Drag the comparison handle to check the edges around hair and shoulders, where segmentation is hardest.

4

Download

Export the finished image. The subject stays sharp while the background carries the blur you set.

At a glance

Model
MediaPipe Selfie Segmentation, run locally in your browser
Effect
Subject stays sharp while the background is blurred — the shallow depth-of-field look of a portrait lens
Adjustable
Blur strength, with a before-and-after comparison slider
Hardest area
Hair and shoulder edges, where segmentation masks are least precise — check these before exporting
Subject type
People. The model is not trained on objects or animals
Privacy
The photo stays on your device throughout

Related tools

Watermark Tool

Add text or logo watermarks with tile mode and drag-to-position controls.

Image Crop Tool

Crop to Instagram, YouTube, circle, passport and custom aspect ratios.

Passport Photo Maker

Resize your photo to standard passport dimensions in seconds.

Social Media Image Resizer

Resize images to exact dimensions for Instagram, Facebook, YouTube and more.

Frequently asked questions

Quick answers about this free online tool.

Last updated

No. All AI processing happens entirely inside your browser using MediaPipe Selfie Segmentation. Your images are never sent to any server, and the segmentation model runs locally on your device using WebGL.

This tool uses Google's MediaPipe Selfie Segmentation model — the same technology used in Google Meet and other video conferencing tools for virtual backgrounds. It runs in real time in your browser via WebGL without any server-side processing.

The AI model needs to be downloaded (~3 MB) on the first use. After that, processing is near-instant and the model stays cached in your browser. All subsequent images process in milliseconds.

Photos with a clear, single human subject produce the best results. The AI works best when the person is clearly visible, well-lit, and positioned in the centre of the frame. Group photos and images with partial visibility may produce less accurate segmentation.

Standard applies a clean Gaussian blur. Portrait adds a slightly darker, desaturated background to mimic DSLR shallow depth-of-field. Bokeh applies a stronger blur with slight contrast enhancement. Soft uses a gentle blur for a barely-noticeable background separation.

Edge Softness feathers the boundary between the subject and the blurred background. A value of 0 gives a sharp, hard edge. Higher values create a smooth, natural transition — like the gradual focus falloff you see in professional portrait photography.

Yes. The preview is scaled down for performance, but the export always renders at your image's full native resolution. The blur radius is also scaled proportionally so the effect looks identical at any size.

If no person is detected (for example, if the image is a landscape or object), the segmentation mask will show no foreground. The output will still be produced — it will just be a uniformly blurred version of the image.