Search requires an extractable text or OCR layer. Pattern detection is assistance, not a guarantee; review every result.
Rasterized output tradeoffs: all original text and OCR become non-searchable; links, forms, bookmarks, accessibility structure, and annotations stop working; signatures become invalid; file size may increase.
Estimated output size appears after loading.
Keep the untouched original securely.
The affected pages are shown for confirmation, and each command is one undoable operation.
Choose a PDF, then drag over text or image areas. Select a mark to move or resize it.
Marks are translucent only for review; exported redactions are always solid. Select a mark and drag it, or drag its lower-right handle to resize.
Before sharing: open this downloaded copy, zoom into every mark, search for every redacted term, and attempt copy-and-paste. Confirm that remaining visible context is safe.
A drawn shape can merely hide content while leaving selectable text underneath. This tool first renders each page, applies solid marks to the pixels, and builds a new image-only PDF. Rasterization is one valid redaction method, though specialist PDF software can also remove selected content objects without rasterizing the rest of a page.
Load the statement, search for the account number if it has selectable text, and draw a separate region over the customer name. Review the two regions on every affected page, apply them, then search and copy from the downloaded copy before sending it.
The new PDF contains full-page images with solid redaction pixels. Original selectable text and OCR, annotations, comments, form fields, links, bookmarks, attachments, metadata, embedded page objects, accessibility structure, and signatures are not copied into the rebuilt output.
All text becomes non-searchable, forms and links stop working, accessibility structure is lost, signatures become invalid, and file size can grow. Higher DPI improves small-print clarity but uses more memory. Retain the untouched original securely.
Yes. There is no signup, payment, watermark, or fixed page limit. Your browser's available memory is the practical limit, so very large PDFs or 300 DPI output may fail on lower-memory devices.
No. The File API reads the PDF into browser memory, processing happens on your device, and the rebuilt file is downloaded locally. The page still makes normal requests for its own HTML, fonts, analytics, ads, and self-hosted tool libraries; it does not send the selected PDF to Starlight Tools.
The marked content is not carried into this tool's rebuilt output, so it cannot be uncovered there. Keep the untouched original secure, and always verify the downloaded copy before sharing.
Yes. You can manually mark image areas on scanned pages. Find and redact only works when the PDF contains an OCR or selectable text layer.
Yes, when the PDF has extractable text. Find and redact lists exact word or phrase matches so you can select individual results or mark all. Pattern presets are assistance, not a guarantee.
The output is a newly created PDF with image-only pages, rather than a modified copy of the original. The completion report checks for inherited metadata, extractable text, and attachments, but you should still inspect the final file before sharing.
No. Rasterized output removes selectable text and OCR layers and does not preserve links, forms, annotations, bookmarks, accessibility structure, attachments, or embedded signatures. Existing digital signatures become invalid because the document is rebuilt.
Every page becomes a full-page image. More pages and higher DPI create more pixels, so 300 DPI output can be substantially larger than the original.
The tool can ask for a password locally when PDF.js supports the document. Some encrypted, permission-restricted, damaged, or unsupported PDFs may still fail; the password is not uploaded or stored.
Open the downloaded copy, inspect every page and mark at high zoom, search for each redacted term, try copy-and-paste, review the sanitization report, and confirm that the remaining visible context is safe.
The local workflow can help with privacy-sensitive documents such as filings, medical records, bank statements, resumes, and public-record releases. It does not guarantee legal or regulatory compliance and is not legal advice; the person releasing the document must validate the result and follow applicable rules.
Publisher: Starlight Robotics · Redaction behavior last verified: 18 July 2026.
The verification matrix covers text PDFs, image-only scanned pages, OCR text layers, manual and searched marks, annotations, form fields, links, document metadata, multiple page sizes, and local image-only export. Checks include opening the output again, requesting its text and attachments, inspecting metadata, and confirming each exported page is rebuilt from rendered pixels.
Known limitations: exact-search geometry can be approximate for unusual fonts, rotated or highly fragmented text; pattern presets can miss or overmatch identifiers; white marks can be visually ambiguous; encrypted or malformed files may not open; very large pages and high DPI can exceed browser memory; rasterized output is not accessible or searchable.
Important: You must validate the final downloaded document before releasing it. This tool does not provide legal advice and does not by itself establish HIPAA, GDPR, court-rule, public-record, or other compliance.
