Blurring your ID yourself vs using a dedicated app
What manual editing gets right, what it misses, and when each approach is enough.
Short answer
- Solid opaque blocks remove the pixels; blur and pixelation are lossy but not always destructive.
- The MRZ is the field people forget — missing it undoes the rest of the redaction.
- The unredacted original usually stays in the camera roll next to the edit, ready to be sent by mistake.
- Manual editing is proportionate for low-stakes sharing where you know exactly where the image stops.
Your phone already has a markup tool. Draw a black box or a blur over your passport number, and it feels done. Mostly, it is — but there are a few specific ways this falls short of what a copy meant to leave your hands should actually protect.
What manual editing gets right
Covering the obvious field — usually the document number — with a black box or a heavy scribble stops a casual glance from reading it. For low-stakes situations where you trust exactly who’s receiving the image and it’s staying with them, that’s often genuinely enough.
Where it falls short
Blur and pixelation aren’t always final. Light blur and low-strength pixelation are lossy, not destructive — under the right conditions, the original content can sometimes be reconstructed. That is not folklore: in Defeating Image Obfuscation with Deep Learning, researchers at Cornell Tech and UT Austin trained neural networks to read faces, objects and digits back out of both mosaiced and blurred images. A solid, opaque block removes the pixel data entirely and doesn’t have this problem; not every markup tool defaults to one.
It’s easy to miss a field you don’t know to look for. Most people manually redacting a passport remember the number. Far fewer think of the machine-readable zone — it re-encodes the number, birth date and more, so missing it undoes the rest of the redaction. Date of birth and signature get missed too, since they don’t look as obviously “sensitive” as a document number.
There’s no watermark. A manually blurred photo is still just a photo — if it’s forwarded beyond where it was meant to go, nothing in the image ties it to the specific request it was made for.
The original often still exists next to it. Markup tools frequently edit a duplicate, but the source photo — unredacted — commonly stays in the camera roll, and it’s easy to grab and send the wrong one under pressure.
Side by side
| Manual markup tool | Dedicated redaction flow | |
|---|---|---|
| Reversibility | Blur and light pixelation are lossy, not always destructive | Solid, opaque blocks by default — the pixels are gone |
| Machine-readable zone | Easy to miss; it does not look “sensitive” | Treated as a standard field to cover |
| Watermark | None | Applied automatically, naming the recipient and date |
| Editable layers | Markup can survive as a removable layer | Flattened on export |
| The original | Unredacted original usually stays in the camera roll | Only the redacted copy is exported |
| Proportionate for | Low-stakes sharing you fully control | Anything leaving your control |
What a dedicated redaction flow does differently
It’s built around what’s actually on a passport or ID: solid, non-reversible boxes by default (not blur), prompts for the fields people forget, a watermark applied automatically, and a flattened export with no editable layer underneath to peel back later. Anonymize my ID does exactly this, entirely on-device — see how to redact a passport or ID card for the full method.
When manual editing is genuinely fine
If you’re sending a partial view to someone who has already seen the original in person, and it never leaves that conversation, a quick manual cover is proportionate. The gaps above matter most exactly when a copy is heading somewhere you don’t fully control — which, for most of the requests in this guide series, is precisely the situation.
Frequently asked questions
Is blurring my ID photo with my phone's markup tool good enough?
It's better than sending an unedited photo, but it has real gaps: blur and pixelation can sometimes be reversed, it's easy to miss a field like the machine-readable zone if you don't know to cover it, and there's no watermark tying the copy to one purpose.
Can blurred text actually be recovered?
Under the right conditions, yes — light blur and pixelation are lossy but not always destructive, and there's documented research on reconstructing text from both. A solid, opaque block removes the pixels entirely and can't be reversed the same way.
What does a dedicated redaction app do differently?
It's built around the specific fields a passport or ID actually has — including ones people forget, like the machine-readable zone — applies solid, non-reversible redaction by default, adds a watermark, and flattens the result so there's no editable layer left underneath.
When is manual editing genuinely fine?
For very low-stakes sharing where you control both ends — say, texting a family member a partial view they've already seen in person. For anything leaving your control, the gaps above start to matter.