WebcamTest.com
← Back to Fix Tools

Real OCR Against Real Blur

Background Anonymizer & Blur Quality Rating

Blur can look soft without actually hiding anything — this tests it directly. One captured frame gets segmented, composited at several real blur strengths, and each version is run through genuine on-device OCR to check whether text is still readable. Both models run entirely on your device.

Camera off

Start the camera, frame your usual background with any readable text or documents visible, then click Capture & Analyze.

Tested Strengths

Result

BlurReadable wordsStatus

Verdict: —

—

Notes

Why blur alone sometimes isn't enough

Light blur softens edges without necessarily destroying the underlying character shapes OCR relies on — especially for large, high-contrast text like signage, whiteboards, or document titles. Genuinely sensitive backgrounds are more reliably protected by solid background replacement than blur, particularly at the lighter blur settings many apps default to for performance reasons.

Reference

Tested Blur Strengths

Blur strengthTypical use
8pxLight blur — many apps' default or low-effort setting.
16pxModerate blur — common mid-range app default.
24pxStrong blur — noticeable but still shows shapes/movement.
32pxHeavy blur — background reduced to soft color blocks.

These are the exact four strengths this tool tests your background against with real OCR.

Step-by-Step

How to Run the Test

  1. Click "Start Camera" and allow access when prompted.
  2. Frame your usual background with any readable text or documents visible.
  3. Click "Capture & Analyze" and wait for both AI models to load and process.
  4. Read the results table showing readable words detected at each blur strength.

Common Questions

Frequently Asked Questions

How is this different from the Background Blur Test?
The Background Blur Test scores edge stability and artifacts — how clean the segmentation itself looks. This tool asks a completely different question: does the resulting blur actually hide readable text, tested with real OCR rather than assumed.
How is this different from the Privacy Leak Scanner?
The Privacy Leak Scanner checks your RAW, unblurred background for readable text before you apply any effect. This tool specifically grades how effective a blur EFFECT is at hiding that text once applied, at several strengths.
Does this send my video or background image anywhere?
No. Both the segmentation model and the OCR engine run entirely on your device — nothing is uploaded at any point.
Why does it only test one captured frame instead of live video?
Running two AI models continuously in real time is a heavy combination without a proven track record in this codebase. A single careful capture-and-analyze pass is more reliable and still gives you a genuine, accurate answer for your actual background.
What should I do if text stays readable at every strength?
Use a solid background replacement instead of blur for that specific background, or physically move sensitive items out of frame — blur has real limits for large, high-contrast text that a stronger blur setting alone may not overcome.