Changelog

recent updates to pdftoolskit.org

27 August 2026 — Word, Excel and PowerPoint to PDF

• Three new converters, and none of them upload your file: Word to PDF, Excel to PDF and PowerPoint to PDF. An Office file is a ZIP of XML, so the archive is unpacked in your browser, the XML is read, and the pages are drawn locally. Every other free Office converter sends your document to a server to be opened by LibreOffice — which is a poor deal when the file is a contract, an offer, a payslip or a pitch deck.
• The output is real text, not a screenshot of a page. The shortcut in a browser is to render to a canvas and paste the bitmap into a PDF: it looks right and fails at everything else — nothing selectable, nothing searchable, nothing for a screen reader, and a file five to ten times larger. A four-page spreadsheet comes out of this at under 10 KB with working Ctrl-F.
• Documents that are not in Latin script are handled honestly. The fonts built into every PDF reader cover Latin-1 and nothing more, so Cyrillic, Greek, CJK and the rest cannot be drawn with them at all. Rather than quietly turning every character into a question mark — which is what happens elsewhere — those documents are detected and rendered as page images with your system's real fonts, and you are told which script triggered it.
• Word: headings at their own style sizes, bold/italic/underline/colour, alignment, bullet and numbered lists with the numbering recomputed, tables with column widths and cell shading, inline images, and the page size and margins the document specifies.
• Excel: all sheets or one, dates converted out of Excel's serial numbers, thousands separators, currencies and percentages read from each cell's own number format, and cached formula results. Columns that do not fit either shrink to the page width or continue on further pages labelled with their column letters — never silently clipped at the margin, which is the usual failure. The header row repeats on every page, and the real header is detected rather than assuming row 1, so a merged report title does not get repeated instead.
• PowerPoint: one page per slide at the deck's own aspect ratio, so 16:9 stays 16:9. Placeholder geometry is inherited from the slide layout and then the master — skipping that chain is why some converters pile every title into the top-left corner. Text that overflows its box is scaled down the way PowerPoint's autofit does, and speaker notes can be appended.
• Fixed while building it: relationship targets written as absolute package paths (/xl/worksheets/sheet1.xml) were being resolved as relative, which made every workbook written by some common tools come out empty.
• The site is now at 59 tools. Legacy .doc, .xls and .ppt are not supported and say so plainly rather than failing obscurely.

27 August 2026 — real encryption, real redaction, 56 tools

• Protect PDF now actually encrypts. The old /lock-unlock page had a "lock" mode that returned an unencrypted copy of your file, because the PDF library it used cannot encrypt. That was misleading and it is gone. In its place is a hand-written implementation of the PDF 2.0 standard security handler — AES-256, revision 6 — running on your browser's own WebCrypto. After saving, the file is re-opened with a second PDF engine to confirm it refuses to open without the password and opens with it, and the result is shown to you.
• Redact PDF now removes the text, not just covers it. A black rectangle drawn over a name leaves the name in the file, selectable and copy-pasteable — the single most repeated document-handling failure there is. Redacted pages are now rasterised so no text object survives, and only those pages: the rest of the document keeps its text layer. The saved file is then re-read and you are told whether any redacted text can still be extracted.
• Redaction works on phones and tablets. It was mouse-only, so on a touch device you could not draw a box at all. Also fixed: dragging a box to the edge of the page used to silently discard it.
• Two pages split out of one. /lock-unlock was trying to serve two unrelated jobs; it is now Protect PDF and Unlock PDF, and the old URL redirects.
• New: PDF → JPG and JPG → PDF, with their own guidance on DPI, page order and when JPG is the wrong choice for a page of text. That takes the site to 56 tools.
• Compress to a target size got more precise. It used to walk a fixed quality ladder and stop at the first setting that fit, often landing far under the target and throwing away quality it did not need to spend. It now binary-searches back towards the last setting that was too big, and reports the DPI and JPEG quality it settled on.
• Split pages in half finds the real gutter. Book scans are never centred exactly, so an exact-50% cut slices through the text. Each page is now measured for its emptiest column near the middle and cut there, with the old exactly-in-the-middle behaviour still available.
• Word count adds sentences, paragraphs, average words per page, estimated reading time, a per-page breakdown and a CSV export. Font checker adds the pages each font appears on and lets you download the embedded font files.
• Fixed: Insert pages was telling search engines it was the homepage. The page was missing its canonical URL, so it pointed at / — which quietly keeps a page out of search results entirely.
• Every tool page now carries a breadcrumb trail, a related-tools block, and a step-by-step section — 22 pages had no how-to at all. Pages are faster too: the 300 KB tool bundle is deferred and the webfont no longer blocks the first paint.

10 May 2026 — initial launch

• 16 tools shipped on day one across four mega-menu categories — Convert, Pages, Edit, Inspect.
• Batch mode by default. Every tool accepts multiple files; one downloadable result per input where it makes sense, or a single combined file when that is the goal (merge, multi-image PDF).
• Convert: PDF → Images (PNG / JPG with adjustable DPI, ZIP per file), Images → PDF (drag-to-reorder, A4 / Letter / Legal / A3 / fit-to-image), PDF → Text, PDF → Markdown (heading inference from font-size jumps), HTML → PDF.
• Pages: Merge, Split (single page / N-page chunks / explicit ranges, ZIP output), Rotate pages, Reorder / delete, Extract pages.
• Edit: Watermark (centered or corner, opacity / rotation / size), Page numbers (six positions, format string with {n} / {N}), Edit metadata (set or strip), Unlock PDF (strip owner-password protections; user passwords cannot be cracked in-browser).
• Inspect: PDF viewer (page navigation, file picker for batches), Metadata inspector (title, author, dates, page count, page size).
• Sample PDFs in datasets for quick tool testing.