Kroma.

Privacy

Files stay on your device

Files stay on your device

Local processing

Your files and typed text are processed in the browser. We do not send document bytes, keyboard contents or clipboard contents to a backend or external AI service. The PDF to Word service does not include OCR.

Storage and cleanup

Files, previews, results and edit history are kept in page memory. Reset or close the page to discard them. Downloaded files remain on your device under your control. Browser caching may retain program assets and fonts, not the tool inputs.

Hosting metadata

The hosting provider may process request metadata, authentication cookies and access logs for the website. This tool does not control the provider’s retention policy. Document contents and editor text are not placed in URLs or analytics. No advertising, payment or analytics integrations are added.

Technical limits and security

Local processing avoids server uploads. Workers isolate heavy tasks; timeouts and file/page/pixel limits mitigate resource use. Browser APIs do not provide hard per-job memory quotas. Malformed or highly compressed PDFs can still stress a device. Files are not rendered as HTML and embedded PDF scripts are not executed.

PDF conversion quality

Word output is editable extracted text with approximate paragraphs and simple tables. Supported images are separate objects. Complex page design, fonts, columns, charts, reading order, forms, annotations and bookmarks may not be preserved. Scanned pages are rejected instead of being passed off as editable text.

Compression

qpdf rewrites structure and eligible images without rasterizing pages. Low is structural lossless optimization; Medium and High allow lossy JPEG conversion of eligible images. Pre-compressed PDFs may not shrink. The unchanged original is offered if no smaller output exists.

Keyboard composition and PDF export

Chinese and Japanese phonetic conversion uses your native operating system IME; the virtual palette is supplementary. Korean virtual keys compose Hangul. PDF export uses browser printing with Unicode shaping and available fonts. Review the print preview before saving.

Libraries

PDF manipulation uses pdf-lib (MIT), PDF parsing uses PDF.js (Apache-2.0), Word generation uses docx (MIT), compression uses qpdf WebAssembly (Apache-2.0; wrapper ISC), and Korean composition uses hangul-js (MIT). Notices are available in /vendor/NOTICES.txt.