Screenshot Redaction Assistant
Cover private details in screenshots, review each redaction and save a flattened image.
The problem it solves
Cover private details in screenshots, review each redaction and save a flattened image.
Who Screenshot Redaction Assistant is for
- People who need a small, purpose-built utility instead of a broad platform.
- Users working with screenshots, redaction, privacy who want the documented v1.3.0 behavior.
- People who prefer an open-source release with visible limitations, source, and license terms.
Intended result
Cover private details in screenshots, review each redaction and save a flattened image.
This summary is reconciled from the current catalog and repository documentation.
Features in v1.3.0
Capabilities below come from the current project README and release documentation.
Draw, move and resize redaction rectangles locally.
Save an unfinished review and reopen it with the matching source image.
Export the final image only after every rectangle has been accepted.
Verified examples
Screenshots are shown only when the current README references a local source image. Otherwise, repository example files are linked directly.
Platforms and implementation
The public release claims only the cataloged platforms and technologies.
Supported platforms
- Windows
- macOS
- Linux
Built with
- Python
- Pillow
What changed in v1.3.0
- Added saved review drafts tied to the source image, with checks that reject the wrong image or an invalid draft.
Changelog
Release notes from the canonical v1.3.0 source tag.
- Reject fractional, boolean and string pixel bounds rather than silently coercing them.
- Download explicit private review drafts containing bounds, selected rectangle and acceptance flags
- Bind drafts to the exact original source image and dimensions, validating all data before replacing review state
- Preserve pending-rectangle final-export guards and warn about unsaved edits
- Reject oversized, malformed and mismatched-image drafts without altering the current review
- Add a local canvas editor with distinct original/preview/final states, guarded normalized recipes, exact union coverage and optional output metadata review.
- Added regression coverage for the audited behavior and invalid inputs.
- Added explicit multi-job and recursive directory batches with reusable rectangle and color recipes.
- Added outlined preview PNGs and terminal review that can keep, skip, replace, or add rectangles before redaction.
- Added optional local OCR suggestions for email, phone-like, IP, and token-like text without silently accepting any suggestion.
- Expanded audit evidence with rectangle identifiers, reasons, origins, dimensions, pixel totals, fill color, review state, and input/output hashes.
- Raised the Pillow runtime floor to the current fixed 12.3 release line after dependency review.
- Released the first complete Screenshot Redaction Assistant command-line workflow.
- Added deterministic reporting, representative examples, cross-platform tests, and explicit limitations.
Quick start
The shortest documented path into the current release.
Before running these source commands, download the v1.3.0 source archive, extract it, and open a terminal in the extracted project folder containing README.md. Follow the tagged README for required runtimes and any platform-specific setup.
python -m pip install .
redact-screenshot examples/sample.json --preview-dir previews --output audit.md
redact-screenshot --input-dir screenshots --output-dir redacted --recipe examples/recipe.json --recursive --reviewInstall .[ocr] plus a local Tesseract executable to use --suggest-ocr. OCR candidates are never silently applied: use --review to keep, skip, replace, or add rectangles. Preview images outline rectangles without flattening them; only final outputs permanently cover pixels.
Current limitations
These boundaries are part of the product and prevent the page from implying unverified capability.
The tool runs locally and does not upload images, include telemetry, or require an account. OCR uses the operator's local Tesseract installation. Redaction is permanent only in the exported flattened PNG. OCR can miss sensitive text or produce false positives; recipes can be wrong for images with different dimensions; rectangle pixel totals can overlap. Operators must inspect every final image and avoid sharing originals or previews.
Privacy and licensing
Review the actual data boundary before using a tool with sensitive inputs.
Privacy and safety
The tool runs locally and does not upload images, include telemetry, or require an account. OCR uses the operator's local Tesseract installation. Redaction is permanent only in the exported flattened PNG. OCR can miss sensitive text or produce false positives; recipes can be wrong for images with different dimensions; rectangle pixel totals can overlap. Operators must inspect every final image and avoid sharing originals or previews.
License and release
Screenshot Redaction Assistant is published under MIT. The current cataloged release is v1.3.0, published 2026-10-17.
Related projects
Related projects are selected deterministically from shared catalog tags, category, and implementation technologies—not popularity or paid placement.
Screenshot Redaction Assistant v1.3.0
Use the tagged release for downloads and release notes. Use the repository for source, issues, contribution guidance, security reporting, and complete documentation.
Page source: current Forge catalog plus README and CHANGELOG from the canonical GitHub release tag. Fingerprint: 32ee7e6b2be33969.