A PDF black box is not the same as a PNG bar

Court PDFs have leaked because someone drew a black rectangle on top of live text. You could still select, copy, paste. That genre is real. Your PNG of a chat is a different file format with a different failure mode.

BBC reporting in January 2019 on a Manafort filing is the public example this post uses: black boxes that still left copy-pasteable text. That is overlay redaction on a document. It is not a screenshot editor. Mixing the two leads people to fear the wrong bug — or to ignore the one they actually have.

What is the difference between an overlay and replaced pixels?

A PDF box is often a drawing instruction. The letters remain in the content stream. A screenshot that you then fill with black in a raster editor replaces samples. There is no text layer to copy unless you later run OCR — and OCR on a solid bar returns nothing useful.

If you screenshot a PDF that was badly overlaid, you may capture the boxes as pixels and still leave the original PDF in the email. The filing is not cleaned because you photographed the page. The PNG might be fine. The attachment list might not.

Does a PNG bar fail the way the Manafort boxes failed?

Not by copy-paste of a text layer. A PNG does not carry a PDF content stream. It can still fail by highlighter tint, by a gap in the bar, by a blur that leaves glyph shape, by sending the original alongside the export. Those are raster failures. BBC’s Manafort story is the overlay failure. Keep the citation in the overlay bucket.

Raster bars: Bars on images. Export format: PNG vs JPEG.

What should you use if your job is legal PDF redaction?

A proper redaction tool that removes content streams. Not a black square in a consumer PDF app that only paints. Not Markup highlighter on a screenshot of page 4. Not “flatten” by printing to a new PDF unless you have verified the text is gone — verification means try to select, search, and copy, then check the file, not the screen.

Do not screenshot a PDF, bar the image, and call the original filing clean. The original filing is the PDF. The screenshot is a picture of a page. Courts and counterparties get the PDF.

What should you use if your job is a Slack snip?

A flat PNG with replaced pixels. You are not producing a court PDF. You are producing an image. Do not paste a PDF into Slack and draw a box in a previewer. Export from an editor that bakes the fill. Inspect the download. Do not assume Slack “flattened” a PDF attachment.

Can OCR bring back a solid bar?

OCR reads glyphs. A uniform rectangle has none. A light smear may have enough shape that a person — or software — can guess a short string. That is the screenshot problem, not the Manafort problem. Solve it with a fill, not with a second PDF box on top of a PNG.

What if someone “prints to PDF” after drawing boxes?

Print-to-PDF can flatten drawings and still keep a text layer, or rasterize the page, depending on the tool. You do not guess. You try select, search, and copy on the output. If text comes back, you have the overlay failure BBC described on the Manafort filing: boxes that look opaque and are not deletion. If the output is a true raster with a solid fill, you have a picture, and the original PDF may still be sitting in the same email.

Two files, two jobs. Redact the document with a tool that removes streams. Redact the screenshot with replaced pixels. Do not use one as a charm against the other. If counsel only received the PNG, the PDF on your disk is still the filing.

If you only needed a chat snip, stay on a raster export and paint there: flatten the bar as PNG in BlurThis.