How it works

Four steps. Then your photo stays yours.

Blurify is a quick editor, not a service. No accounts, no archives, no analytics on the photos themselves. Here's exactly what happens between drop and download.

  1. Drop or pick your images

    Up to 20 images per batch, 20 MB each, JPEG / PNG / WebP / HEIC. Files go directly from your browser to a private S3 bucket via a one-time upload URL. Nothing passes through our server when uploading.

  2. We highlight the sensitive bits

    Auto-detection finds faces, license plates, document text with PII patterns, screens, monitors, ID cards, and other common privacy leaks. Whatever's found is shown as an outlined region — nothing is blurred until you say so.

  3. Adjust the result yourself

    Toggle any suggestion off. Drag the image to draw a region we missed. Per-region: pick blur, pixelate, solid fill, or color overlay, adjust strength, choose a color. Live preview right on the canvas. Your edits are the final word.

  4. Export. We forget.

    Hit download — you get the result back as JPEG, WebP, or PNG with EXIF and GPS metadata stripped. The original and the export are deleted within an hour, automatically. We never look at them.

What we don't do. No accounts to sign into. No images saved to a gallery. No tracking pixels on the editor. No third-party analytics on your photos. No background ML training on what you upload.

Where the detection actually comes from

Different categories use different vendors behind a single interface so we can pick the right tool for each job:

  • Faces — AWS Rekognition `DetectFaces` (cheapest competent face detector, 27 landmarks per face).
  • License plates — Google Cloud Vision `OBJECT_LOCALIZATION`. We filter for the "Vehicle registration plate" class and refine the bbox.
  • Document text & PII — Google Cloud Vision `DOCUMENT_TEXT_DETECTION`, then a local regex pack flags emails, phone numbers, IBANs, US/UK SSN/NI, and Luhn-valid card numbers.
  • Screens, monitors, phones, cards — Google Cloud Vision named objects with bounding boxes.

None of these vendors retain your images. We're a thin pipeline that calls them, drops the result on your canvas, and walks away.

Ready to try it?

Open the editor →