Every PDF job, done on your device.

Paragraphs and headings carry over as editable text. Complex layouts, tables and images are simplified — that is the honest trade of a text-flow conversion. Scanned pages have no text to carry (OCR is coming).

Anyone opening the file will need this password. There is no recovery — keep it safe.

English, via the Tesseract engine running in your browser. A searchable PDF keeps every page pixel-identical and lays recognized text invisibly on top, so you can select and search it.

Publish this form as a link. People fill it in their browser and their answers are encrypted to your key before anything is uploaded — Bindery cannot read them.

Free includes 1 open form link with up to 25 responses. Pro: unlimited links, up to 1,000 responses each.

Tokens: {n} row number, {file} the form’s name, {Column name} any column — e.g. {file}-{Last name}. Checkboxes tick on yes / true / 1 / x; dropdowns and radios match an option by name.

Looks for what someone could still pull out of this file: lines only in earlier saved versions, text and images no page uses, selectable text under black boxes, images or annotations or in hidden layers, invisible, tiny and off-page text, hidden form values, thumbnails, bookmarks, metadata, attachments, scripts and redaction marks never applied. Snippets are masked here and in the report.

Mixed page sizes become uniform: content is scaled proportionally and centered. Text stays real text.

Honest scope: Bindery re-parses the file — leniently if it must — drops unreadable objects and rewrites a clean structure. That fixes broken cross-reference tables and truncated trailers. A file with no recoverable page tree can’t be saved, and Bindery will say so.

Embedded JPEGs come out byte-for-byte untouched. Other decodable bitmaps are saved as lossless PNG. Masks and exotic formats (CMYK, indexed, 1-bit) are skipped and counted honestly.

Lines are reassembled in reading order, page by page. Scanned pages have no text layer — run OCR first, then extract.

Flattening makes a PDF final: nothing left to edit, clear, or accidentally reveal in an annotation. Highlights drawn by Bindery’s Markup tool are already flat.

Cleaned pages are rebuilt as page images, so any text layer is lost — run OCR afterwards to make the clean scan searchable. Deskew detects rotation up to ±3°.

Honest scope: this is a real PAdES-style digital signature (SHA-256 with RSA or ECDSA, certificate chain embedded). It proves the file hasn’t changed since you signed it. With the device-generated key the certificate is self-signed, so readers mark the identity as untrusted; import a certificate from an authority your readers trust and the same signature shows as trusted. No server sees the document either way.

How it works: the file is sealed on this device with a fresh AES-256 key. Only the sealed package is stored — Bindery cannot read the document or even its name. The key is the part of the link after the #, which browsers never send to any server, so whoever you send the link to is the only one who can open it. The package is deleted when the link expires or the download limit is reached. Creating links needs a Bindery account (it is the abuse control); opening them does not.

What conversion does: embeds an sRGB output intent, writes XMP with the PDF/A-2b identification, sets a document ID, removes JavaScript, embedded files and multimedia annotations, generates form appearances and drops annotations that have none. It is a practical preflight, not a certified validator — the report says so and points to veraPDF for a formal verdict.

Writes trim, bleed and art boxes, an sRGB output intent with an embedded ICC profile, the PDF/X-4 identification in Info and XMP, and Trapped. Removes JavaScript and embedded files. PDF/X-4 allows RGB and transparency; a CMYK-only PDF/X-1a needs colour conversion in the print workflow.

Read like a screen reader would: tagging, language, title, alternative text on figures, a text layer on every page, headings, form-field descriptions, link text, bookmarks and table headers. Colour contrast and the quality of alt text still need a human. Nothing is changed; the file is only read.

Auto-tag builds a first structure tree for an untagged file: headings by size, paragraphs, figures with placeholder alt text, reading order top to bottom. Tables and lists are not detected.

Each signature is checked on this device: the signed bytes against the digest, the signature against the embedded certificate (RSA, RSA-PSS or ECDSA), the certificate’s validity dates, whether the signature covers the whole file, and any RFC 3161 timestamp. Bindery has no certificate-authority store, so it reports who issued the certificate rather than claiming the identity is trusted — that judgement is yours or your reader’s.

Honest scope: this re-flows the document — headings, paragraphs, bullet lists, simple tables and inline PNG/JPG images, on the page size and margins Word specified. Fonts become Helvetica; columns, text boxes, headers/footers and tracked changes are simplified. For a pixel-exact rendering, export from Word itself.

Each page becomes one slide, rendered as a sharp image at 1600 px and sized to the page’s own aspect ratio. Opens in PowerPoint, Keynote and Google Slides. The text on the slides is part of the image — that is the honest trade of a faithful conversion.

Honest scope: every visible sheet becomes a clean grid — column widths from the workbook, the first row treated as a header and repeated on each page, numbers right-aligned, dates formatted, formulas shown as their last calculated value. Charts, merged-cell layouts, colours and conditional formatting are not reproduced.

Honest scope: slides are re-drawn at their real size — backgrounds, filled shapes, text with sizes, bold, italics, alignment and bullets (positions inherited from the layout and master), pictures and tables. Fonts become Helvetica; gradients are flattened to one colour; shadows, SmartArt, charts, embedded media and animations are not reproduced, and the result says what was skipped.

One sheet per page. Rows come from lines of text; columns are detected from the gaps between them, so ruled tables and aligned columns carry over well, while free-form prose lands in a single column. Plain numbers become real numeric cells; scanned pages have no text layer — run OCR first.