PDF Magic Exporter

Tools for rendering accessible books as PDFs and creating searchable, structured derivatives of scanned PDFs.

Web toolbench

The dependency-free web interface keeps Google Books export, OCR, local AI review, and compression in separate reusable modules. Each panel validates its own inputs and builds the corresponding local CLI command.

npm run serve

Open http://localhost:4173 in a browser. Run the frontend command-builder tests with npm test.

Google Play Books export

export-google-book.py captures the visible pages available through a Google Play Books reader session and combines them into a PDF. It uses a persistent browser profile so the user can sign in and retain reader access.

Install the Python dependencies and Playwright browser:

python3 -m pip install img2pdf playwright
python3 -m playwright install chromium

Export a volume by its Google Books ID:

python3 export-google-book.py BOOK_ID

By default, captures are written to BOOK_ID-pages/ and the combined document to BOOK_ID.pdf. The exporter resumes from existing captures, advances until the reader stops, and removes the final reader-completion capture before PDF assembly.

OCR, compression, and structure tagging

The lossless PDF workflow can create searchable OCR, identify heading levels, apply PDF structure tags, recompress compatible image streams without changing decoded pixels, and optionally incorporate a local Ollama review. See the OCR workflow for requirements and commands.

Enhanced PDFs record Author (/Author, dc:creator) from recognizable signatures or large letterhead names, then existing author metadata, then the local user's full name or username. Creator (/Creator, xmp:CreatorTool) retains the original PDF Info and XMP values separately, even if they disagree or resemble an author name. Producer (/Producer, pdf:Producer) preserves the original value for each metadata format and appends ; PDF Magic Enhancer once; if missing, it is PDF Magic Enhancer. Name detection uses the searchable text and is heuristic.

The OCR and compress-and-tag enhancement workflows accept optional --title, --author, --description, --keywords (alias --tags), --publisher, --rights, --language, and --identifier values. Unspecified descriptive fields retain source metadata rather than being invented. PDF Info Title, Subject, and Keywords map to XMP dc:title, dc:description, and dc:subject/pdf:Keywords, respectively. Publisher, rights, language, and identifier are mapped into Dublin Core XMP, and language also sets PDF /Lang. The web toolbench offers editable metadata fields.

Pass --source-url URL to ocr-scanned-pdf.sh to store the original document URL in the XMP pdfmagic:href property. Existing original URLs are retained when processing an enhanced local copy again.

SEC concept-release comment references

download-sec-comments.py downloads the PDF and raw HTML submissions listed for an SEC file number. Public-comment PDFs, raw HTML comments, and meeting memoranda are stored in separate directories.

Set an identifying SEC User-Agent locally and provide the file number:

export SEC_USER_AGENT='Your Name your@email.com'
python3 download-sec-comments.py S7-27-15

If SEC_USER_AGENT is unset, the tool prompts for it in an interactive shell. Use --dry-run to inspect the files and classifications without downloading.

DO NOT USE FOR PRORESED RULE, CP ONLY

This tool is intended only to refresh local reference copies.