pdf-ocr

Appearance

Privacy ​

Two separate things are described here, and they have different answers: the application you run, and this documentation site.

The application ​

pdf-ocr is software you run yourself. There is no hosted service, no account, and no telemetry of any kind. Nobody but you sees the documents you convert.

Concretely, the application:

  • sends nothing anywhere. There is no analytics, no crash reporting, no update check, and no outbound network call. The web page loads no third-party assets — Tailwind is vendored into the repository, and the Content-Security-Policy is same-origin.
  • stores your files on your own disk, under UPLOAD_FOLDER. The uploaded PDF is deleted as soon as the conversion ends, whether it succeeded, failed or was cancelled. The converted file stays until you download and discard it, or until cleanup removes it — files older than 24 hours are deleted, and task records an hour after their last update.
  • stores no personal data of its own. The session cookie holds a list of conversion ids you started, nothing else. There is no identity to store.
  • writes logs containing filenames, chosen options and errors. They go to stdout by default, and to a file only if you set LOG_FILE. They are yours.

One caveat worth being explicit about: some OCR engines download language models on first use. EasyOCR and PaddleOCR fetch these from their own project infrastructure. That is a download of model data, not an upload — your documents are never involved — but it is an outbound connection, and it is theirs rather than ours. Tesseract, the default, does not do this: its language packs come from your system package manager.

This documentation site ​

These pages are static files served by GitHub Pages. There is no analytics script, no cookie set by this site, and no tracking of any kind in the pages themselves.

GitHub serves them, and as the host it processes request data — including IP addresses — for delivery and abuse prevention. That is governed by GitHub's Privacy Statement, not by this project. If that matters to you, the entire site can be read from the repository or built locally.

The site's search runs entirely in your browser; queries are not sent anywhere.

Contact ​

For anything about this project, including privacy questions, open an issue or email fabrizio.salmi@gmail.com. For a suspected vulnerability, use the private channel described on the Security page instead.