Pixelation or AI Erasure of Sensitive Data

Why hide data in photos

Marketplace listings, reports, stories, and portfolios often accidentally contain personal or business data: license plates, document numbers, on-screen messages, ticket QR codes, addresses on signs. Removing them in a heavy editor takes time; this tool is built specifically for quickly preparing an image for publication.

Pixelation mode

Selected areas are turned into large colored blocks. The data becomes unreadable, but the area remains visibly censored. This is handy when your audience or regulations require visible anonymization (internal reports, news, compliance).

The pixel size slider controls coarseness: too fine a pixel can sometimes leave readable outlines of digits, so for license plates it is better to use a medium or high value.

AI Erasure mode (LaMa)

The LaMa neural network (Large Mask Inpainting) completely removes the fragment and fills the hole with background: a bumper instead of a plate, a wall instead of a sign, fabric instead of a badge. The publication looks like an ordinary product or reportage photo without a pixelated patch.

The rectangle mask is expanded and slightly rounded before being sent to the server so the edges do not look like a cut-out box. After successful erasure the canvas is updated, so you can immediately process the next area on the already clean image.

Keep in mind: LaMa invents texture rather than revealing what was under the plate. For privacy, that is a desirable property.

Typical scenarios

• Cars and marketplaces. Remove or pixelate the license plate before publishing a listing.
• Real estate. Addresses, intercoms, names on doorbells, personal photos on shelves in the frame.
• Office and products. Screens with CRM, employee badges, stickers with internal codes.
• Documents in frame. Work papers on a desk that accidentally got into the shot (not for forging documents, only for cleaning up the background of a publication).

How to use

1. Upload an image.
2. Select all sensitive areas with a small margin around the edges.
3. Choose pixelation (explicit censorship) or LaMa (natural-looking image).
4. Apply the processing and zoom in on the preview: make sure the digits/text are truly unreadable.
5. Download the PNG and only then publish.

Tips for reliable hiding

• It is better to mask a bit more than to leave the edge of a plate digit visible.
• For glossy plates and reflections, LaMa often looks more natural than pixels.
• If a hint of an outline remains after LaMa, add another area on top and repeat.
• Do not rely on fine pixelation for small on-screen text; increase the pixel size.

Responsibility

The tool is intended for lawful preparation of your own or permitted materials for publication. Do not use it to hide evidence or forge documents. The user is responsible for the content and distribution of the result.