Talk PDF

Read less. Listen more. Turn any PDF into an audiobook, right in the browser.

Run it

No build step. Serve the folder with any static server, then open it in Chrome, Edge, or Safari:

cd Talk-PDF
python3 -m http.server 8080
# open http://localhost:8080

Opening index.html straight from disk also works in most browsers. A local server is more reliable, especially for the optional OCR fallback.

Files

• index.html: markup for the landing page, reader, players, and dialogs
• css/styles.css: design tokens (dark and Paper themes), layout, and responsive rules
• js/app.js: application logic, organized into modules:
  • PDFProcessor: extracts text and page structure with pdf.js
  • OCRProcessor: lazy-loaded tesseract.js fallback for scanned PDFs
  • TextProcessor: detects headings, paragraphs, and sentences, and removes running headers and footers
  • TTSProvider / WebSpeechProvider: pluggable text-to-speech layer
  • Narrator: playback engine (chunking, seeking, speed, voice, and position)
  • UI modules: uploader, document viewer, contents, highlighter, search, voice selector, speed controller, player, mini player, full-screen player, and settings

Adding a TTS provider

Extend TTSProvider and implement init, getVoices, speak(text, opts), cancel, isSpeaking, and prepare(nextText). Audio-file providers should prefetch the next chunk in prepare. Then pass your provider to new Narrator(provider).

Notes

• PDFs are processed locally. Nothing is uploaded to a server.
• Voices come from the browser and operating system through the Web Speech API. Edge "Natural" voices and Chrome's Google voices sound best.
• Elapsed and remaining times are estimated from an average reading pace.
• Preferences and reading positions are saved in localStorage.