Hide Details

Hide Details is a Raycast extension that detects sensitive text and faces in clipboard images so you can redact them before sharing. Review what it found or copy a redacted image with a single command. All processing happens locally on your Mac using Apple Vision.

Commands

Command	Description
Redact Clipboard Image	Scan the clipboard image, review the detected details, and copy the redacted image.
Redact and Copy	Scan and redact the clipboard image immediately, then copy the result without opening a view.

How to use

1. Copy a screenshot or image to your clipboard.
2. Run Redact Clipboard Image in Raycast.
3. Check the redacted image preview and detections, then select Copy Redacted Image from any row.
4. Paste the result wherever you need it.

For a quicker workflow, assign a hotkey to Redact and Copy in Raycast. Copy an image, press the hotkey, then paste.

If you copy something else while Redact and Copy is processing, it preserves the newer clipboard and asks you to run the command again. Copying from the review view is an explicit action and replaces the current clipboard.

Use Scan Current Clipboard Again to retry with your saved preferences. Use Scan Current Clipboard with Accurate OCR for difficult text. These actions read the image currently on your clipboard and replace the previous preview. The Accurate OCR action applies to that scan without changing your saved Text Recognition preference.

Review results

The Redacted Image row shows the output and the number of distinct masked regions. Detection labels mask the recognized text so it stays hidden while screen sharing. Each detection row shows the same image and offers Copy Redacted Image. One OCR line can match more than one category. These appear as separate detections but share one masked region.

The OCR and Face percentages describe recognition confidence. They do not measure whether every sensitive detail was found. Turn off Show OCR Confidence in extension preferences to hide the OCR percentages. Custom word match identifies a literal rule from Always Hide These Words.

A result with no detections can still be copied, but it has no added masks. Check the preview before sharing. If copying fails, the preview remains available so you can retry Copy Redacted Image.

Copying saves readable clipboard data in memory for recovery if the image write fails. Unavailable clipboard formats do not block copying the finished image. If a failed write cannot be fully recovered, the error asks you to copy the original again.

What it detects

• Email addresses and phone numbers.
• Card numbers that pass a Luhn checksum.
• Common secret formats, including Stripe keys, GitHub tokens, AWS access key IDs, Slack tokens, and JWT-shaped strings.
• IPv4-shaped addresses.
• Personal names and words you specify.
• Text matching your custom regex.
• Faces.

Text detections redact the full line recognized by OCR. Automatic detection can miss details, so check the resulting image before sharing it.

Card and phone detection checks numbers separately from neighboring numeric details. This includes cards beside expiry dates, phones beside ticket numbers, adjacent phone numbers, and phones beside IPv4 addresses. Other long numeric identifiers can resemble phone numbers and cause extra lines to be masked.

Preferences

Text Recognition defaults to Fast for clear screenshots. Fast recognition can miss details in small or difficult text; choose Accurate for these images. Accurate recognition can take longer on its first run while macOS prepares its models.

Preference	Description	Default
Text Recognition	Choose Fast or Accurate OCR.	Fast
OCR Confidence	Show OCR confidence percentages in the detection list.	On
Redaction Style	Choose Blackout, Pixelate, or Smooth Blur.	Blackout
What to Hide	Comma-separated categories: email, phone, card, secret, ip, name, face.	All categories
Always Hide These Words	Comma-separated names or strings to detect as names. Keep the name category enabled to use this list.	Empty
Custom Regex	Optional regex applied to each OCR line, independently of the built-in categories.	Empty
Padding	Extra pixels around each detected region, from 0 to 1000.	4

Use Blackout when sensitive content must be unreadable. Blur and pixelation can retain visible information.

Blackout is the default when no style has been saved. An existing Pixelate or Smooth Blur preference stays selected until you change it in Open Extension Preferences.

Unknown category names and invalid padding stop processing with an error. Category names are case-insensitive. Padding accepts fractional values as well as whole pixels within the allowed range.

Custom rules

Use Always Hide These Words for names or literal strings. For patterns such as invoice numbers, internal IDs, or file paths, set Custom Regex in the extension preferences.

Literal word matching is case-insensitive, ignores whitespace, and matches substrings within an OCR line. Entries shorter than two characters are ignored. Keep name in What to Hide to enable these rules.

Enter a pattern directly, without JavaScript-style / delimiters:

\bINV-\d{4,6}\b

Combine rules with |. Use (?i) for case-insensitive matching:

(?i)\bINV-\d{4,6}\b|confidential|client\s+id

The pattern uses Apple's regular expression syntax. It runs against each original OCR line, including spaces. For example, Project\s*Nebula also matches a space introduced by OCR between those words. A match redacts the entire line and appears as Custom Regex in the detection list.

Custom Regex applies independently of What to Hide. Clear that category field to use only your regex. Leave Custom Regex empty to disable it. Matches that contain no characters are ignored. Invalid patterns or patterns that take too long stop redaction with an error, leaving the clipboard unchanged.

Privacy

Images are processed on your Mac without uploads. Clipboard images are read directly into memory. The extension does not write unredacted clipboard images to disk.

Each scan owns a redacted output image in a private temporary directory. The immediate command removes it after copying or a copy failure. The review command removes it when the result is replaced or the view closes. Output files abandoned by a stopped process are removed after 24 hours on a later run. Live review results are preserved. Temporary files from older versions use a 24-hour cleanup threshold.

The output is a flattened PNG with no editable layers.

Troubleshooting

Issue	What to do
Clipboard has no image	Copy the image itself, then choose Scan Current Clipboard Again. A text-only clipboard cannot be redacted.
Clipboard changed while scanning	The scan or automatic copy stopped. Run the command again for the image currently on your clipboard.
Small or difficult text was missed	Use Scan Current Clipboard with Accurate OCR and inspect the new preview. The first Accurate scan can take longer while macOS prepares its models.
Unknown category or invalid padding	Open extension preferences. Use the category names listed above and a padding value from 0 to 1000.
Custom Regex is invalid or takes too long	Correct or simplify the pattern in extension preferences, then scan again.
Preview expired or output could not be read	Scan the current clipboard again to create a new result.

Development

Requires macOS 14 or newer, Raycast, Node.js 22.22.2 or newer, and Xcode with Swift 6.3 or newer for the Swift build plugins.

Install dependencies and start development from the project directory:

npm ci
npm run dev

Native code lives in the Swift package at swift/hide-details. TypeScript calls its @raycast functions through swift:../swift/hide-details imports, using the same integration as Color Picker. Raycast compiles and packages the native code during development and builds. Generated native assets and build caches are ignored by Git.

To validate and build the extension:

npm run lint
npm run build
npm run typecheck
npm test

Run npm run publish to build the extension and start Raycast's publishing flow. No manual native build or checked-in executable is required.

npm test builds the extension and its Swift bindings before running the regression suite. The tests cover mask pixels and crop direction, Fast and Accurate OCR, dotted cards with only card detection enabled, cards and phones beside other numbers, masked review labels, face detection, small text, preference validation, and scan cleanup. Native clipboard tests use an isolated pasteboard and check PNG/TIFF input, copied output, restoration after rejected writes, and protection against stale copies without replacing your clipboard.

License

MIT