Scripts and tools for preparing documents of various formats for NotebookLM. Handles PDF, DJVU, EPUB, FB2, MOBI, DOC/DOCX, and many other formats -- converting, OCR-ing, and splitting them into word-count-limited chunks suitable for upload.

The main entry point. Converts and splits documents of various formats into word-count-limited chunks ready for NotebookLM.

./prepare_all.sh [input_dir]

input_dir defaults to the current directory. The script finds all supported files, converts/OCRs as needed, and splits them into chunks in the output directory.

# Process all documents in the current directory
./prepare_all.sh

# Process a specific directory
./prepare_all.sh /path/to/books

# Customize chunk size and output
MAX_CHUNK_WORDS=300000 OUTPUT_DIR=chunks ./prepare_all.sh /path/to/books

# Set OCR language and keep intermediate PDFs
OCR_LANG=rus+eng KEEP_PDF=1 ./prepare_all.sh /path/to/books

# Skip OCR entirely
SKIP_OCR=1 ./prepare_all.sh /path/to/books

• MAX_CHUNK_WORDS - Maximum words per output chunk (default: 400000)
• OUTPUT_DIR - Directory for output files (default: out)
• OUTPUT_PREFIX - Prefix added to output filenames (default: none)
• KEEP_PDF - Set to 1 to keep intermediate PDFs from format conversions (default: 0)
• OCR_LANG - Tesseract language(s) for OCR (default: rus+eng). Supports + syntax for multiple languages.
• SKIP_OCR - Set to 1 to disable OCR for all formats (default: 0)

prepare_all.sh relies on scripts from the subdirectories below, plus:

• pdftotext, pdfinfo, qpdf - PDF manipulation (poppler-utils, qpdf)
• ocrmypdf - PDF OCR
• ebook-convert - Format conversion (install Calibre)

For DJVU support: ddjvu, djvutxt, djvused, tesseract (djvulibre, tesseract-ocr).

Scripts for working with PDF files -- splitting, combining, OCR-ing.

• ocr_pdf.sh - Adds OCR text layer to image-only PDFs in place, with progress bar
• split_pdf.sh - Splits large PDFs into word-count-limited chunks
• combine_pdf.sh - Combines multiple PDFs into chunks by word count
• pdf_process.sh - All-in-one: split, combine, and generate a manifest

Scripts for extracting text from DJVU files and optionally converting to PDF.

• ocr_djvu.sh - OCRs DJVU files with ocrodjvu, embeds text layer in place, extracts .txt
• djvu_to_txt.sh - Extracts text from DJVU via page-by-page tesseract (does not modify originals)

Scripts for working with text and ebook formats.

• split_text.sh - Splits large text files into word-count-limited chunks
• split_epub.py - Splits EPUB files into smaller parts
• epub-to-text.py / epub-to-text2.py - Converts EPUB to plain text

Browser automation for NotebookLM.

• add_links_script.py - Automates adding links as sources to a NotebookLM notebook via Playwright

# Fedora
sudo dnf install poppler-utils qpdf ocrmypdf djvulibre tesseract tesseract-langpack-rus calibre

# Debian/Ubuntu
sudo apt install poppler-utils qpdf ocrmypdf djvulibre-bin tesseract-ocr tesseract-ocr-rus calibre

# macOS
brew install poppler qpdf ocrmypdf djvulibre tesseract calibre

pip install -r text-tools/requirements.txt

For browser automation, also run:

playwright install chromium