Skip to content

PDF2Bard Companion Guide

AudioBard natively ingests clean EPUB and TXT files. If your manuscript or public domain book is in PDF format, raw text extraction usually results in mangled audio due to layout artifacts.

For this reason, we created PDF2Bard (oscarbol09/pdf2bard) — a dedicated, rule-based PDF pre-processor built specifically for audiobook TTS pipelines.

The Problem with Raw PDF Text Extraction

When traditional tools extract text from a PDF: - Visual Line Breaks: Sentences are split mid-clause at margin boundaries, creating unnatural pauses in TTS speech. - Marginal Hyphenation: Words like dis- / cover are read as two separate broken words ("dis" ... "cover"). - Running Headers & Footers: Page numbers and chapter titles printed on every page are read aloud by the narrator. - Damaged Dialog Marks: Guillemets and em-dashes get mangled, confusing the character attribution engine.

What PDF2Bard Solves

  1. Smart Paragraph Reflow: Distinguishes genuine paragraph/dialog breaks from soft visual wraps.
  2. Automatic De-Hyphenation: Seamlessly welds hyphenated words across line boundaries without breaking compound words like "well-known".
  3. Header & Footer Stripping: Detects repetitive running titles and page counters and eliminates them.
  4. Dialogue Punctuation Normalization: Standardizes quotes (“ ”, « », —) so AudioBard can accurately extract characters and detect speaker turns.

Quick Usage

Visit the repository: oscarbol09/pdf2bard

# Clone & install PDF2Bard
git clone https://github.com/oscarbol09/pdf2bard.git
cd pdf2bard
pip install -e .

# Convert PDF to clean EPUB
pdf2bard convert novel.pdf --output novel.epub

# Ingest directly into AudioBard
audiobard generate novel.epub -o audiobook.mp3