Audiobooks are the format most self-publishers skip and the one with the least crowded shelf. They are also genuinely more work than an ebook, and the decision is worth making on arithmetic rather than enthusiasm.
The arithmetic first
Length is set by the manuscript, not by choice: about 9,300 words per finished hour at normal narration speed. So a 60,000-word book is roughly six and a half hours of audio.
| You do | Effort per finished hour | Money |
|---|---|---|
| Narrate and edit yourself | 6–8 hours | Equipment only |
| Narrate, pay an editor | 2 hours | Per finished hour |
| Hire a narrator | ~1 hour of admin | Per finished hour, the largest cost |
| Synthetic voice | Under an hour | Small or nothing |
Six and a half finished hours narrated yourself is therefore forty to fifty hours of work. That is the number people do not price in, and it is the reason most self-narrated audiobooks are abandoned in chapter four.
Which books justify it
- Fiction and memoir: yes, usually. Performance is part of the product and listeners notice a good reading.
- Practical non-fiction: often not. It is consumed in fragments, and tables, diagrams and code do not survive being read aloud at all.
- Anything visual: no. If the interior design is carrying meaning, audio loses it.
- **A book by someone with an audience who wants *them***: yes, and narrate it yourself even if you are imperfect. The voice is the point.
Narrating it yourself
Cheaper in money, expensive in hours, and better than people expect if you are willing to be boring about the setup.
The room matters more than the microphone. A mid-range USB microphone in a wardrobe full of clothes beats an expensive one in a kitchen. You are trying to eliminate reflections; soft surfaces do that and money does not.
- One room, one position, one set of levels. Never change any of it mid-book; the join is audible.
- Record in chapter-sized sessions. Your voice changes over a day.
- Punch and roll: when you fumble a line, stop, back up a sentence, carry on. Fixing it in the edit is slower.
- Leave a few seconds of room tone at the start of every file. The editor needs it.
- Water, no dairy, and a gap after eating. Mouth noise is most of what editing removes.
Expect to hate your own voice for the first hour and to stop noticing after that.
Hiring a narrator
The cost is per finished hour and it is the largest line in this whole undertaking, usually more than editing and cover combined.
- Audition with a real passage, including dialogue and anything technically awkward. Never judge from a showreel.
- Agree the pronunciation list up front: names, places, jargon.
- Listen to the first fifteen minutes before they record the rest. This is the single most important step and the one most often skipped.
- Royalty-share arrangements exist on some platforms: no money up front, a share of sales forever. Cheap if the book does not sell, expensive if it does.
Synthetic narration
Now accepted by the major retailers, including Amazon's own virtual-voice programme, and it has moved from unusable to adequate.
Honestly: it is obviously synthetic to a listener, it handles emphasis and irony poorly, and it is fine for reference-style non-fiction that people dip into. It is a poor fit for anything where voice carries feeling.
It is also the only option that makes a short book worth producing at all, because the fixed cost is near zero. The stores require disclosure; follow it.
What the stores want
Broadly consistent across ACX, Findaway Voices and the rest:
| Requirement | Typical |
|---|---|
| Format | WAV or high-bitrate MP3 |
| Sample rate | 44.1 kHz |
| One file | Per chapter |
| Opening and closing credits | Required, as their own files |
| Room tone | A few seconds at head and tail |
| Loudness and noise floor | Within a stated range |
The loudness and noise-floor specs are where submissions fail, and they are a mastering problem rather than a performance one. Any competent audio editor hits them; it is not something to eyeball.
Retail sample: pick a passage that is interesting rather than the first page, which is usually front matter.
Where the text comes in
The narrator, human or synthetic, reads a script, and your manuscript is not one yet. It needs a pass that:
- Removes "as the table below shows" and everything else that assumes eyes.
- Spells out what a listener cannot see.
- Handles front and back matter, most of which is not read aloud.
- Marks pronunciation for anything ambiguous.
That pass is much easier from a plain-text source than from a laid-out file, because you are editing words rather than a document. It is the same argument as everywhere else in production: bukpress builds the PDF and the EPUB from one Markdown manuscript, and that manuscript is also the cleanest thing to turn into a narration script: one source, three outputs, instead of three documents drifting apart. How to self-publish a book is where this sits in the sequence.
Frequently asked questions
How long does it take to record an audiobook? Six to eight hours of work per finished hour if you narrate and edit yourself, so forty to fifty hours for a typical 60,000-word book. A hired narrator reduces your part to auditions and review.
Can I use AI to narrate my audiobook? Yes. The major retailers accept synthetic narration and require it to be disclosed. It is adequate for reference non-fiction and noticeably flat for fiction.
Do I need a recording studio? No. A quiet room with soft furnishings and a decent USB microphone clears the specs. The room matters far more than the microphone.
Is an audiobook worth it for a self-published author? For fiction, memoir and author-led non-fiction, often. For practical non-fiction with tables and diagrams, usually not, because the format loses too much.
Can I record it in sections over months? Yes, if the room, position and levels never change. Inconsistency between sessions is the most common reason a submission is rejected.