Installation

Docker

The Docker image contains pd3f, its web app and everything OCR needs. It is published for amd64 and arm64, so it also runs natively on Apple Silicon.

docker run -p 127.0.0.1:8000:8000 -v pd3f:/data pd3f/pd3f:1.0

Or with Docker Compose: save docker-compose.yaml in a folder of its own and run docker compose up there. Jobs and the language models then live in the folder pd3f-data next to it.

Then open http://localhost:8000. The image is a download of about 250 MB, about 700 MB unpacked. Its Tesseract has the data for German, English, French, Italian and Spanish; for OCR in another language, build an image on it that adds that language’s package, such as tesseract-ocr-rus. The web app has no authentication: keep it on localhost, or put a reverse proxy with access control in front, see In production with nginx.

Python

pip install pd3f          # library and command line tool
pip install "pd3f[web]"   # with the web app

pd3f needs Python 3.11 or later. OCR also needs Tesseract with the data of the document’s language and unpaper, e.g. apt install tesseract-ocr tesseract-ocr-deu unpaper. Without Ghostscript, OCRmyPDF warns about each JPEG image it leaves as it is; the output is the same. PDFs that have text need none of them.