Is there any mention or guide about using ebook2audiobook to preprocess an epub's text for TTS inpu | ElevenLabs | Page 1

hallow oysterBOT Mar 5, 2026, 8:42 AM

#

AI-generated Answer for sjallard_bsr

Short answer: No.

I can’t find any official ElevenLabs documentation or help‑center article that mentions or recommends using the third‑party “ebook2audiobook” tool for preprocessing EPUB text.

Official guidance is to import EPUBs directly into ElevenLabs:

Studio supports EPUB (and PDF, DOCX, TXT, HTML/URL) and provides structure-aware import; use Heading 1 for chapter titles so chapters are auto-detected. (help.elevenlabs.io)
For publishing, supported manuscript formats include EPUB, PDF, TXT, DOCX, and HTML; text may be normalized for audio compatibility. (help.elevenlabs.io)
ElevenReader/Publishing workflows also state you can import your book as an EPUB, DOCX, or PDF. (elevenlabs.io)

Sources:
Which file formats can I import with Studio? – ElevenLabs
What formats are supported for publication? – ElevenLabs
Create your audiobook with ElevenReader Publishing

stark star Mar 5, 2026, 8:44 AM

#

But there are no bult-in tools, or discussed tools, to convert numbers to words, abbreviations, etc ?

hallow oysterBOT Mar 5, 2026, 8:44 AM

#

AI-generated Answer for sjallard_bsr

ElevenLabs does have built‑in and documented options for converting numbers, dates, and similar items to spoken words (text normalization), plus tools to control acronyms/abbreviations:

Built‑in normalization in the UI: When you generate TTS on the website, normalization runs automatically to improve pronunciation of numbers, dates, etc. The best‑practices guide also notes normalization is enabled by default across TTS models. (elevenlabs.io)
API control: Use the apply_text_normalization parameter (‘auto’, ‘on’, ‘off’) on TTS endpoints (including Convert and streaming/timestamps). This spells out numbers/dates when enabled and can add latency. Model restrictions apply for some v2.5 models (Flash/Turbo). (elevenlabs.io)
Changelog and Help Center details: Normalization via apply_text_normalization was added in Oct 2024; for v2.5 models it may be limited (e.g., only ‘auto’/‘off’ unless on Enterprise). The help article also explains defaults and Studio settings. (elevenlabs.io)
Project/Studio workflows: Project creation (Studio/Audio Native) also exposes apply_text_normalization and adds an ‘apply_english’ mode. (elevenlabs.io)
Language‑specific normalization: apply_language_text_normalization (currently Japanese only) can be enabled, with higher latency trade‑offs. (elevenlabs.io)
Abbreviations/acronyms and custom pronunciations: Use Pronunciation Dictionaries (.PLS/TXT) with alias or phoneme entries (e.g., map “UN” → “United Nations”) across Studio, Dubbing Studio, and the Speech Synthesis API. (elevenlabs.io)

If you need deterministic expansion (e.g., always expand units/URLs), the best‑practices page also recommends preprocessing (e.g., regex/LLM prompts) alongside the built‑in normalization. (elevenlabs.io)

stark star Mar 5, 2026, 9:15 AM

#

Built-in normalization is a disaster for FR-ch numbers (Swiss French)

#Is there any mention or guide about using ebook2audiobook to preprocess an epub's text for TTS inpu