Evidence-backed Windows capture guide
How to blur or redact sensitive information in a screenshot
Remove private information from a screenshot with a deliberate choice between presentation blur and permanent black redaction, then verify the exported pixels.
Crop unrelated content first. Use blur or pixelation only when the goal is visual de-emphasis; use an opaque black redaction when the value must be removed. Export a new image, reopen it, zoom in, and confirm the sensitive pixels are gone before sharing.
Check release statusKeyword focus: blur sensitive information screenshotReduce exposure before editing
Capture the smallest useful region and close private notifications. Cropping a password manager, email sidebar, or account menu completely out of the image is safer than covering it later. Keep an untouched original in a protected location when the screenshot is evidence.
Distinguish blur from redaction
Blur and pixelation preserve a transformed version of the underlying visual structure, which is useful for presentation but can be a poor confidentiality boundary. Permanent black redaction replaces the selected pixels with a solid value in the exported bitmap. An annotation rectangle that can be moved later is not the same as a flattened redaction.
Use automatic detection as a review aid
The frozen build can locate sensitive email regions locally, but the published fixture deliberately demonstrates the limit: the email is covered while a phone number remains visible for manual review. Automatic detection must never be described as complete privacy verification.
- Crop unrelated content
- Run local sensitive-region detection if useful
- Review every suggested region
- Add manual redactions
- Flatten to a separate output
- Reopen and inspect every private value
Verify the final pixels
Do not rely on the editor canvas alone. Open the exported image in another viewer, zoom in, inspect the edges of every box, and confirm that copyable text, metadata, filenames, and nearby context do not reveal the value indirectly.
First-party workflow evidence
Controlled input → settings → expected → actual
- Controlled input
- A controlled privacy fixture containing an email address and a phone number.
- Mode and settings
- Frozen 0.6 local auto-redaction workflow, permanent black output, separate destination file.
- Expected result
- The recognized email region is permanently changed without overwriting the input; unmatched content remains visible for manual review.
- Actual 0.6.0 result
- The published before/after artifacts show the email replaced while the phone number remains visible, accurately demonstrating both capability and limitation.

Inspect all 78 named tests · Verify this guide workflow · Verify artifact hashes
Failure modes and limits
Automatic detection can miss phone numbers, names, IDs, addresses, faces, or unusual text. Blur and pixelation may not meet a policy that requires irreversible removal.
Stop when the capture omits required context, a scroll seam repeats or skips content, a recording stream cannot be decoded, OCR changes a critical value, or a redaction leaves private information visible.
Recovery and verification
Return to the untouched original or .uvcap project, add or resize redactions, export a new file, and repeat independent visual inspection.
Record the application version, capture target, pointer and delay settings, scroll axis or audio choice, output format, and final artifact hash when the result must be reproducible.
Bound to one frozen build
These statements use the 293,888-byte executable with SHA-256 3832C20012F30B07C2D61ACECCAFC9949B4738F065C53D23D2C781D6BCCF5DB7. The product audit passed; public signing did not.
Frequently asked questions
Is blur the same as redaction?+
No. Blur is a presentation treatment; opaque flattened redaction is the safer choice when pixels must be removed.
Does automatic detection find every private value?+
No. The public test intentionally shows an email detected while a phone number still needs manual review.
Does the app overwrite the original?+
No. The tested workflow writes a separate output.
Official and primary sources
Links were checked August 15, 2026. External claims are limited to the cited Microsoft source.
UtiliVera Capture
Capture deliberately. Verify the output. Share only what belongs.
The public download opens after trusted code signing; until then, these pages preserve evidence without starting the 30-day clock.
See the product →