Blur vs. Pixelate vs. Blackout: Which Redaction Method Should You Use?
Last reviewed: August 18, 2026
Quick answer
Use solid blackout for passwords, API keys, identity numbers, financial details, and anything that must not remain visually recoverable. Blur and pixelation are useful when you want to reduce recognition while preserving the context of an image, but both transform the original pixels instead of replacing them completely.
This guide documents the three methods as they are currently implemented in PrivaCanvas. It is not a promise that automatic face detection or any single editing method makes an image legally anonymous.
| Method | What PrivaCanvas does | Good fit | Main limitation |
|---|---|---|---|
| Blur | Shrinks the selected region and scales it back with smoothing | Casual face or background obscuring | Shapes and color patterns can remain |
| Pixelate | Shrinks the region and scales it back without smoothing | Visible, deliberate masking that keeps layout context | Large blocks still preserve coarse structure |
| Blackout | Replaces every output pixel inside the selected rectangle with a solid dark color | Secrets, credentials, identifiers, and high-risk text | An incomplete selection can still expose information around the edges |
How the three methods differ
Blur keeps a softened version of the image
PrivaCanvas downsamples the selected area and enlarges it again with smoothing. Fine detail is removed, but the result still contains transformed information from the original region. That makes blur visually gentle, not a good default for an API key, password, document number, or other exact secret.
Pixelation keeps coarse blocks
Pixelation also downsamples the selected region, then enlarges it without smoothing. The result makes editing obvious and can work well for casual face privacy. It still preserves block-level colors and shapes, so it should not be treated as deletion of the underlying information.
Blackout replaces the selected pixels
The blackout mode paints a solid dark rectangle over the selected region in the exported canvas. For high-risk text and identifiers, this is the strongest of the three PrivaCanvas options. The important remaining risk is selection: if even one character or edge falls outside the rectangle, it may still be visible.
A simple decision rule
- Would exposure cause account, identity, financial, employment, or legal harm? Use blackout and inspect every edge.
- Is the goal only to make a person or background detail less recognizable while keeping visual context? Blur or pixelation may be appropriate after review.
- Did automatic face detection create the masks? Zoom in and check every face. Detection can miss a face or mark the wrong area.
- Are you sharing a screenshot with credentials or private text? Add masks manually; face detection does not look for text.
Repeatable safety check
You can test a redaction method without using anyone's private photo:
- Create a synthetic image containing a made-up name, a fake API key, and a simple face drawing.
- Open the image in PrivaCanvas, draw one mask, and download a copy in each mode.
- Zoom into the downloaded PNG and inspect the full border of every mask.
- If text secrecy matters, run your own offline OCR test against the downloaded copies. PrivaCanvas does not upload the file or run text recognition.
- Record the browser, date, image dimensions, mask size, and result so the test can be repeated after an editor update.
The exported PNG is flattened: the mask is rendered into the new image rather than stored as a movable layer. Your original file is unchanged on your device, so make sure you share the redacted download rather than the original.
Privacy and implementation notes
- Image decoding, face detection, editing, and export run in the browser.
- The current editor does not send images, filenames, face locations, or mask coordinates to Cloudflare Web Analytics.
- Automatic face detection is a review aid, not identity recognition and not a completeness guarantee.
- The editor generates a new PNG rather than copying the source file's EXIF metadata into the download.
See the Privacy Policy and Third-party Licenses for the current implementation record.
Frequently asked questions
Can blurred or pixelated information be recovered?
They retain a transformed version of the source pixels, so you should not rely on them as irreversible deletion of a secret. Whether a specific detail remains recognizable depends on the image, selection size, method, and later processing.
Is a black rectangle always safe?
In the flattened PrivaCanvas download, pixels inside the selected rectangle are replaced. A rectangle that is too small can still leave information visible outside its edges, and the unedited original remains wherever you stored it.
Which method should I use for faces?
Blur or pixelation can be suitable for ordinary visual privacy. Use blackout when the consequences of recognition are high, and always inspect automatic detection results before sharing.
Which method should I use for API keys and passwords?
Use blackout with generous padding around the complete secret. Then inspect the downloaded file—not only the editor preview—before sharing it.