Privacy · 8 min read

Blur Faces and Licence Plates Before You Post a Photo: How and Why

By the · Published

A photo of a street festival for the club website, a picture of a damaged car for a forum, a screenshot of a conversation for a complaint, a class photo for the family chat: everyday photos often show more people and details than the one thing they are meant to show. Once posted, they can be copied, searched and recognised by face recognition services, and they cannot be called back.

This guide explains when it is worth hiding faces, licence plates and documents, what the law says in broad terms, which way of covering is really safe, and how to do it before the photo leaves your phone or computer. The examples use Toolula's Blur Faces & Plates, which detects and covers areas in your browser.

When anonymising makes sense

The rule of thumb: whoever is in the picture by chance, and is not the subject of it, should not be recognisable when the picture is published. That includes passers-by in the background, other people's children, other guests at a private event, and the cars parked in front of your house when you sell it. Children deserve special care, because they cannot decide for themselves and photos follow them for years.

Licence plates are a less obvious case. A plate identifies a car and, through the vehicle register, its keeper; together with the place and time of the photo it reveals where someone was. On a photo for a car sale, an accident description or a parking complaint, the plates of other cars have no business being readable, and many sellers hide their own plate as well.

Documents are the third category: a screenshot showing names and phone numbers, a photo of a letter, a parcel label in the background of a product photo, a whiteboard in an office photo. What is readable on the screen is readable for everyone who sees the post.

What the law says in broad terms

In Germany, the Art Copyright Act (Kunsturhebergesetz, KUG) protects the right to one's own image: pictures of people may generally be published only with their consent (§ 22 KUG). There are exceptions in § 23 KUG, for example for people who appear only as an accessory next to a landscape or place, for gatherings such as demonstrations, and for contemporary history, but they are narrower than many people assume and always require weighing the person's interests.

The GDPR also treats photos of identifiable people as personal data. Purely personal or household activity, such as sharing pictures with family, is outside its scope. A post for a company, a club website or a public social media account generally is not, and then a legal basis is needed for every identifiable person shown. In other countries the rules differ in detail, but the direction is similar.

Covering faces and plates does not replace consent where it is needed, but it removes the problem for everyone who is only in the picture by chance. This section is a general overview, not legal advice.

Which cover style is really safe

Not every blur hides what it should. A light blur over a face may still leave the shape of the head, the hair and the clothing recognisable, which is enough for acquaintances. A light blur over text can sometimes be partly reconstructed, especially for large, known fonts such as those on licence plates. Mosaic with large blocks is safer than a soft blur, and solid black is the only cover that leaves nothing to reconstruct.

Toolula offers Solid black, Strong blur and Mosaic. For text, documents and licence plates use solid black. For faces, a strong blur or mosaic usually looks less intrusive and is safe enough as long as the blocks or the blur are large compared to the face.

Equally important is how the cover is stored. A box drawn in an editor as a separate layer, or an annotation in a PDF, can sometimes be removed again. Toolula paints the cover into the pixels of a new image, so there is nothing underneath.

Step by step

The detectors work on your device. The face detector also finds small faces in the background of group photos, but faces turned away, in profile or half hidden can be missed; the plate detector finds number plates of cars and motorbikes, and Document blur reads the words on a screenshot or document with on-device text recognition and boxes every word it is reasonably sure of. Detection is a starting point, not a guarantee, which is why you check every box before anything is covered.

  • Drop the photo, screenshot or image of a document.
  • Choose what to find: Pixelate faces, Censor licence plates or Document blur, then click Protect now.
  • Check every box under Edit boxes: delete boxes over things that may stay visible, move or resize the others.
  • Switch to Draw boxes and cover anything the detection missed, such as a name badge or a house number.
  • Pick the cover style, confirm with Yes, protect and click Download safely.

The details people forget

Reflections in windows, mirrors and sunglasses show faces and screens. Name badges, jersey numbers, school logos on uniforms and distinctive tattoos can identify a person as reliably as a face. A house number, a street sign or a view out of a window can reveal where someone lives.

The file itself can give away more than the picture. Photos from a phone usually carry the GPS position where they were taken, the date and time and the device model in their EXIF data. The download from Blur Faces & Plates is a new picture that does not copy the original's EXIF data. For photos you don't need to blur, Remove EXIF removes these details without re-encoding the image; the guide Remove Metadata Before Sharing explains what they contain.

Keep the original. Once the blurred version is posted, you may want to use the photo differently later, and anonymising is not reversible.

Videos

The same applies to videos, with the added difficulty that people and cars move. Blur Video Areas finds faces and plates in a video and follows them from frame to frame, and the Video Privacy Check also removes the location and device data that phones write into video files. The guide on removing location data from videos describes that workflow.

Screenshots of chats and documents

Screenshots are shared far more casually than photos, yet they often contain more personal data: names, profile pictures, phone numbers, email addresses, order numbers and timestamps. Before posting a screenshot of a conversation, a delivery notice or an error message, ask what it needs to show and cover everything else.

Crop first. Cutting away the status bar, the contact list and unrelated messages removes most of the problem before anything has to be covered. Then use Document blur, which reads the words with on-device text recognition and puts a box on each of them; delete the boxes over the words that should stay readable, such as the error message itself, and keep the rest covered. Solid black is the right style for text, because a blurred line of text in a known font can sometimes be guessed.

Frequently asked questions

Is a blurred face really unrecognisable?

Only if the blur is strong. A light blur can leave a person recognisable for people who know them. Use strong blur or mosaic for faces and solid black for text and plates.

Can a blurred licence plate be read again?

A weak blur over large, standard characters can sometimes be partly reconstructed. Solid black leaves nothing to recover, so use it for plates and documents.

Do I still need consent if I blur faces?

For people who are recognisably the subject of a picture, covering their face changes little. For people only in the picture by chance, making them unrecognisable avoids the question. This is not legal advice.

Is my photo uploaded for detection?

No. Face, plate and text detection run in your browser, and the covered image is created on your device.

Does it work on screenshots and scanned documents?

Yes. Document blur recognises the words on the image and boxes each one. Delete the boxes of words that may stay visible and use solid black for the rest.

How this guide is kept current: it is checked against the official sources it cites and against the Toolula tools it describes, and revised when a rule, deadline or tool changes. The date above shows the last revision.

Open Blur Faces & Plates