What You Can Import Into an AI Audiobook Tool

August 7, 2026

AudioProducer.ai imports two things: an .epub file, which arrives already broken into chapters, or text you paste in chapter by chapter. That is the entire list. There is no .docx, .txt, .pdf or .mobi import, so if you write in Word, Scrivener or Google Docs the job is exporting an EPUB or pasting. This guide covers how to get an EPUB out of the tool you write in, when pasting is simply better, and what survives the trip.

The short answer: EPUB or pasted text

AudioProducer.ai takes two kinds of input. You can upload an .epub file, and the project arrives already broken into chapters, with chapter titles and body text carried across as a starting point. Or you can start a blank project and paste your text in, chapter by chapter. That is the entire list. There is no .docx import, no .txt, no .pdf, no .mobi.

Leading with the limit changes every other question on this page. If you write in Word, Scrivener, Obsidian, or on a serial platform, none of those is an import format for us. Each one is a place you export from. So the useful question is not whether we support Scrivener. It is whether you should get an EPUB out of Scrivener, or whether the piece is short enough that pasting it takes less time than converting it.

Everything past that point is the same regardless of where the words came from. Once the text is in a project, you assign a voice per character, run Auto-Assign to tag lines by speaker, add music and soundscapes, and render a chapter to one MP3 with a single click. The import step is the only place your original format matters at all.

Getting an EPUB out of the tool you actually write in

Most writing software can produce an EPUB, and in most cases it is one menu away rather than a project you have to schedule.

Scrivener compiles to EPUB directly, and its compile settings are also where you decide what counts as a chapter, which matters more than it sounds. Ulysses, Campfire, Dabble, and Novlr all export EPUB from their share or export menus. Atticus, Vellum, and Reedsy are book formatting tools, so EPUB is their main output and the file they hand you is usually the cleanest one you will get.

Word and Google Docs do not export EPUB natively. Both export to DOCX, and there is a conversion step between DOCX and EPUB that you run yourself before you arrive. Apple Pages does export EPUB, from the Export To menu. Markdown from Obsidian or Notion converts to EPUB cleanly because there is so little formatting to lose, which is the reason plain-text workflows tend to have the least trouble here.

One thing to check after any conversion: open the EPUB and confirm the chapter breaks are where you think they are. A converter that produced one enormous chapter has technically succeeded and practically failed, because the chapter is the unit we generate and export.

When pasting is simply better than converting

Conversion is worth it for a finished manuscript with thirty chapters. It is not worth it for a short story, a single episode, or a draft you are still rewriting weekly.

Pasting into a blank project is the right route in several common situations. Short work is the obvious one: a short story or a single chapter is faster to paste than to package. Serials are the second, and possibly the bigger one. If you post a chapter a week on Royal Road, Wattpad, Scribble Hub, or Kindle Vella, you are never going to have a finished EPUB, and waiting for one means never starting. Paste each chapter as you publish it and the audio version keeps pace with the written one.

Work in progress is the third case. If the text is still moving, an export is a snapshot that is wrong within a week. Pasting the current version of a chapter into the project costs a few seconds and always reflects what you have now.

There is also a quiet advantage to pasting that nobody mentions: you clean the text on the way in. Pasting is manual enough that you notice the stray page header, the "Chapter 12" line you do not want read aloud, the footnote marker that would be narrated as a number. An EPUB import brings those through untouched.

What survives the trip, and what does not

Narration is spoken words in order. Anything that is not spoken words in order has to be dealt with, and it is worth knowing which of your formatting matters before you import.

Italics, bold, and font choices carry no meaning into audio. Where you used italics for emphasis, the narrator's delivery has to carry it instead, which is what the per-line dialogue emotion controls are for. Where you used italics for internal thought, decide how you want that read and tag those lines to the character whose thought it is.

Footnotes and endnotes are the most common problem. In print a reader chooses to look down. In audio there is no down, so a footnote either interrupts the sentence or disappears. Decide per note: fold the content into the sentence, or cut it. Non-fiction writers hit this hardest, which is why white papers and research papers need more preparation than a novel of the same length.

Images, charts, and tables do not convert. A table of numbers read aloud is not information, it is noise. Where the visual carries something the listener needs, the usual answer is to describe the takeaway in one sentence in the text and ship the visual separately as a companion PDF.

Front matter is worth a deliberate decision rather than a default. Copyright pages, dedications, and epigraphs all get read if you leave them in a chapter. We have a Book Intro that reads once before the first chapter and a Chapter Intro that reads before every chapter, so title and author announcements have a home that is not your manuscript text.

Chapters are the unit, so split before you import

One-button generation renders a whole chapter into one audio file, and each chapter downloads as its own MP3. There is no per-line export and no bulk exporter that flattens everything into a single file. That makes the chapter the unit of everything: generation, review, download, and any re-render you do after a fix.

The practical consequence is that a manuscript imported as one giant chapter is painful to work with, because a typo on page 300 means re-rendering the entire book. Splitting it properly is the difference between a five minute fix and a long one. If your chapter breaks are unclear in the source, the chapter splitting guide covers where to put them and how long a listening chapter should run.

Serial writers get this for free, since each posted episode is already a chapter. Non-fiction writers with long sections often benefit from splitting more finely than the print book does, because a forty minute listening block is a lot to ask.

The specific case of a PDF

PDF is the worst source format you can start from, and we would rather say so than let you find out one chapter in.

A PDF describes where marks sit on a page. It does not reliably describe reading order. Copying text out of a two column layout can interleave the columns. Headers, footers, and page numbers land in the middle of sentences. Hyphenated line breaks turn into broken words. Scanned PDFs have no text at all until they have been through OCR, and OCR errors are silent, so the first sign of trouble is often a narrator confidently reading a word that was never in your book.

If a PDF is genuinely all you have, budget real cleanup time. Extract the text, read it through once looking for interleaving and stray page furniture, fix the hyphenation, and only then paste it in chapter by chapter. If you also have the file the PDF was made from, use that instead. It is almost always faster.

Every import route, by where your writing lives

Below is every source we have written a specific guide for, grouped by where the text is coming from. Each guide covers the export step for that tool or platform and the preparation that source needs before narration.

Writing and drafting apps

These are the tools you draft in. All of them either export EPUB directly or produce text clean enough to paste.

  • Scrivener, where the compile settings also decide your chapter breaks
  • Ulysses, exporting a sheet group as one book
  • Campfire, taking the manuscript out of a worldbuilding project
  • Dabble, exporting a manuscript with its scene structure intact
  • Novlr, from browser draft to narrated chapters
  • Sudowrite, and what to check in an AI assisted draft first
  • A ChatGPT story, including the cleanup that chat output needs
  • Obsidian notes, where Markdown converts unusually cleanly
  • A Notion doc, and how to get the blocks out in reading order

Book formatting tools

If your book has already been through one of these, you are one export away from the cleanest EPUB you will get.

Document formats

The files themselves, including the two we accept and the ones that need a conversion step first.

Web serials, fanfiction, and story platforms

Serial writers rarely have a finished file, so almost everything here is a paste as you publish workflow.

Newsletters, blogs, and posts

Short recurring pieces, where the archive is the manuscript and pasting is nearly always the right call.

Scripts, research, and other source material

Sources that are not novels, where the preparation work is usually about structure rather than format.

Where to start

If you have an EPUB, upload it and you are past the only step this page is about. If you do not, decide between converting and pasting on length rather than principle: a finished book earns the conversion, a serial or a short piece does not. Either way, the next decision is narration, which is covered in how to direct AI audiobook narration, and the full path from manuscript to finished file is in how to make an audiobook with AI. If you would rather look around first, every guide on this blog is listed on one page.

The free account gives you 1,200 words a month with no credit card, which is enough to import a chapter and hear what your book sounds like before you commit to converting anything. Paid plans start from $39.99 a month.

Frequently asked questions

Can I upload a Word document or a PDF?
No. AudioProducer.ai accepts an EPUB upload or text pasted into a blank project, and nothing else. There is no docx, txt, pdf or mobi import. If your manuscript is in Word, either convert it to EPUB first or paste it in chapter by chapter. For a PDF, converting is usually not worth it: extract the text, clean up the page headers and hyphenation, then paste.
Do I keep my chapter structure if I paste instead of importing an EPUB?
You create the chapters yourself when you paste, one chapter per paste. An EPUB import brings the existing chapter structure and titles across automatically. Either way the chapter is the unit that matters: one-button generation renders a whole chapter to one MP3, and each chapter downloads as its own file, so it is worth splitting the text properly before you generate.
What happens to images, footnotes and tables in my manuscript?
None of them narrate. Footnotes either interrupt the sentence or vanish, so decide per note whether to fold the content into the text or cut it. Tables and charts read aloud as noise, so state the takeaway in one sentence and ship the visual separately as a companion PDF. Italics and bold carry no meaning into audio either, so emphasis has to come from the narration itself.

Related posts

← Back to Blog