Transcribing your book sessions: software, a human service, or yourself
For most book projects, automatic transcription software is the right default: it costs little, and the transcript is raw material you will rewrite anyway. Pay for a human service when the audio is poor, the voices overlap or the words must be exact. Type it yourself only for a few short files. Whatever you choose, read the provider's terms on data use before uploading client stories.
The three options side by side
Prices below were taken from each provider's own pricing page on October 20, 2026. We name two well-known providers as examples of each type, not as recommendations; many others exist.
| Automatic (software) | Human service | Do it yourself | |
|---|---|---|---|
| How it works | You upload audio or record inside the app; speech recognition produces text | A trained person listens and types, often starting from a machine draft | You listen and type, or correct a machine draft yourself |
| Accuracy | Good on clear audio with one or two speakers; weaker on names, jargon, accents and crosstalk | Rev states its human transcription is "99%+ accurate" | As good as your patience |
| Your cleanup time | Check names, numbers and quotes against the audio | Light: spot checks | All of it |
| Example prices (Oct 20, 2026) | Otter: free Basic plan with 300 transcription minutes a month; Pro $16.99 per user a month billed monthly, or $8.33 a month billed annually. Rev: free plan with 45 AI minutes a month; Essentials $29.99 per seat a month, or $305.90 billed annually | Rev: $1.99 a minute | Free, plus your hours |
| Who hears your audio | The company's systems; check the terms for training and human review | At least one person outside your team | Only you |
| Best for | Most book sessions | Poor audio, heavy accents, group sessions, exact quotes | A few short files, or highly confidential material |
Prices and plan limits from otter.ai/pricing and rev.com/pricing, checked October 20, 2026. Otter's page also showed lower promotional prices on the day we checked; Otter's Pro plan listed 1,200 in-app recording minutes and 10 audio or video file imports a month, which matters if you record on a separate device and upload. Check current terms before you buy.
What it costs for a whole book
Cost depends on hours of audio, so start with your session plan. If you record one 60 to 90 minute session per chapter for a 12-chapter book, plus three extra sessions, you will have somewhere around 15 to 22 hours of audio. Here is the arithmetic at the human rate listed above:
| Hours of audio | Minutes | Human transcription at $1.99 a minute |
|---|---|---|
| 10 hours | 600 | $1,194 |
| 15 hours | 900 | $1,791 |
| 20 hours | 1,200 | $2,388 |
Rate from rev.com/pricing, October 20, 2026, before any subscription discount. Hours are examples.
The same 20 hours, 1,200 minutes, would sit within a single month's allowance on some software plans listed above, if you transcribe as you go. The money difference between software and humans is real; the bigger difference for most authors is how much checking they want to do themselves.
Accuracy: what actually goes wrong
For book work you do not need a perfect transcript. You need one where you can find the stories and trust the details. Software tends to fail in predictable places, so check these against the audio:
- Names of people, companies and products. Many tools let you add a custom vocabulary; add your key names before your first session.
- Numbers: dollar amounts, percentages, years. A misheard "fifteen" for "fifty" can change a story.
- Jargon from your field.
- Who said what when two people talk over each other. Separate audio tracks help a lot (see the recording setup guide).
- Exact quotes you plan to print. Always check those word for word.
A practical routine: transcribe each session the day you record it. Read the transcript within a week, while you still remember what you meant, and fix names and numbers as you go. Mark the strongest stories with a highlight. That pass doubles as the first step of editing.
Cleanup time
Cleanup is not the same as editing. Cleanup makes the transcript faithful to what was said. Editing turns what was said into a chapter. Keep them separate:
- Software transcripts: fix names, numbers and speaker labels, and mark anything you could not hear. Do not fix grammar yet; you will rewrite those sentences anyway.
- Human transcripts: spot check a few passages against the audio and move on.
- Do it yourself: type what you need, not everything. If you know a ten-minute stretch is a tangent, write "[tangent about hiring, 14:20 to 24:10]" and skip it.
Ask for "clean verbatim" if a human service offers the choice. It drops the ums and false starts while keeping your words. Full verbatim, which keeps every stutter, is meant for legal and research work and makes book editing harder.
Privacy and client stories
Book sessions are full of other people: clients, colleagues, family. Before you upload a recording anywhere, decide what is in it and who should hear it.
- Read the data use terms first. Look for whether the provider may use your recordings or transcripts to train its systems, whether people at the company or its contractors may review them, how long files are kept, and how to delete them. For example, Otter's privacy policy (effective June 16, 2026) lists "training our proprietary AI technology on de-identified audio recordings" as a purpose. Rev's privacy policy (updated October 14, 2026) says it does not cover information Rev processes on behalf of its customers, so for uploaded files you need to read Rev's customer terms as well. Policies change; read the current version.
- Check your own obligations. If you are a lawyer, doctor, therapist, financial advisor or anyone else bound by confidentiality rules or client contracts, those rules apply to recordings too. When in doubt, ask whoever handles compliance for you.
- Disguise as you speak. Change names and identifying details on tape ("a client, I will call her Dana, who ran a regional bank"). Then nothing identifying ever leaves your hands.
- Get permission for any story you will publish with real names, and keep it in writing.
- Delete when done. Once the chapter is final and backed up in your own storage, delete the files from services you no longer need.
Which to choose
- Choose software if your audio is clear, you are one or two speakers, and you are willing to check names and numbers.
- Choose a human service if your audio is rough, there are three or more voices, you have a strong accent the software struggles with, or you need exact quotes and do not want to check them yourself.
- Do it yourself if you have only a few hours of audio, the material is too confidential to share, or you want to relive the sessions as a way into editing.
- Mix them. Software for everything, then a human service for the one session recorded in a noisy restaurant.
If you plan to work with a professional writer, ask how they handle transcription before you pay for any. Interview-based ghostwriting services usually record and transcribe the sessions themselves; at Scribe Media, for example, a writer interviews you over a series of calls and writes the book from them.
Next: turn the transcript into a chapter. The transcript to chapter guide walks through the five edits with a worked example.
Common questions
Is automatic transcription accurate enough for a book?
Usually, yes, because a book transcript is raw material that you will rewrite anyway. Check names, numbers, technical terms and anything you plan to quote word for word against the audio. If the audio is poor or there are several voices talking over each other, a human service is worth the extra cost.
How much does human transcription cost?
Rev listed human transcription at $1.99 a minute on its pricing page on October 20, 2026. At that rate, 15 hours of recordings (900 minutes) would cost $1,791. Prices change, so check the provider's own page before you order.
Do transcription services use my recordings to train AI?
Some say they may. Otter's privacy policy, effective June 16, 2026, lists training its AI on de-identified audio recordings as one of its purposes. Policies differ and change, so read the privacy policy and terms of any service before you upload client stories or confidential material.
Should I transcribe my own recordings?
Only if you have very few hours of audio or cannot share the recordings with anyone. Typing out speech is slow. A middle path is to run software first and then correct the transcript yourself while listening, which also helps you start editing.
Sources: Otter pricing and Otter privacy policy (effective June 16, 2026); Rev pricing and Rev privacy policy (updated October 14, 2026); scribemedia.com/services. All checked October 20, 2026. Published by Scribe Media.