How to transcribe an interview for your book: the steps, the styles and a template

This guide is the hands-on part: you have an interview recording (you being interviewed, or you interviewing someone for the book) and need a transcript you can trust. Choosing between software, a human service and doing it yourself is covered in the transcription options guide. Turning the finished transcript into prose is covered in the transcript to chapter guide. Everything in between is here.

The nine steps

1. Get a clean recording and give it a name

   Before you transcribe anything, listen to the first and last minute of the file to make sure the whole interview is there and both voices are audible. Then rename the file so it sorts in book order, using the same pattern you will use for the transcript, for example ch04_first-hire_s1_2026-10-21.m4a and ch04_first-hire_s1_2026-10-21_transcript.docx. The recording setup guide has the full naming system.

2. Choose the transcript style

   Decide before you start whether this is clean verbatim (the default for book work) or full verbatim (for passages you will quote exactly or keep as a record). Switching halfway produces a transcript that is neither. The styles are compared in the next section.

3. Do a first pass with a tool, or by hand

   For most people the first pass is software: upload the file and let it produce a draft. If the material is too confidential to upload, or you only have a short clip, type it yourself. Either way, the first pass is about getting words on the page, not getting them right.

4. Correct it while you listen

   Play the audio and read along, fixing what the first pass got wrong. Slow the playback down if you can. Do not rewrite sentences to make them read better; that is editing, and it comes later. The only goal of this pass is that the transcript matches the recording.

5. Add speaker labels

   Use a short, consistent label for each person, set in bold or capitals at the start of each turn: INTERVIEWER and AUTHOR, or first names. Start a new paragraph every time the speaker changes. Software often splits one person into two speakers or merges two people into one, especially when voices overlap, so check every change of speaker.

6. Add timestamps

   Put a timestamp in hours, minutes and seconds (00:12:31) at the start of each speaker turn, or at least every few minutes. Timestamps are how you get from a line in the transcript back to the moment in the audio when you need to check a quote or hear the tone of a story.

7. Mark what you cannot hear

   Never guess silently. Write [inaudible 00:12:31] for words you cannot make out, and [crosstalk 00:20:05] where two people talk at once. If you can make a reasonable guess, write it with a question mark: [Halvorsen? 00:14:02]. Brackets tell your future self, and anyone else who reads the transcript, exactly which words are not confirmed.

8. Build a proper names list

   Keep a short list at the top of the transcript, or in a separate file per book, of every person, company, product, place and term that appears, spelled correctly. Ask the speaker to confirm spellings you are not sure of. Most transcription tools let you add custom vocabulary; add the list there too, and the next session's first pass will be better.

9. Save and back up

   Save the corrected transcript next to the recording, under the matching file name. Keep one copy on your computer and one somewhere else, such as cloud storage or an external drive. Keep the original audio until the book is published: it is the only way to settle a question about what someone actually said.

Full verbatim, clean verbatim and edited transcripts compared

Transcription providers use slightly different names for the same things, so here is what each term means. Rev, a transcription provider, describes clean verbatim as a transcript that has been "lightly edited for easy readability" and says it never paraphrases; full verbatim (Rev calls it regular verbatim) writes down every word, including fillers, false starts, stuttering and interjections from other speakers (Rev). Some providers use "intelligent verbatim" for the same thing as clean verbatim (University Transcription Services), while others use "intelligent" or "edited" for a transcript that paraphrases or summarizes (SpeakWrite). When you order a transcript or set up a tool, check which one you are getting.

Which one a book author wants

Clean verbatim, for most book work. You will rewrite the material into chapters, so the ums and false starts are noise you would only delete later. Clean verbatim still keeps the speaker's exact words, which matters because those words are where the voice of the book comes from. An edited transcript that paraphrases loses that voice, and you cannot quote from it.

Full verbatim, only when exact speech matters. Use it for a passage you plan to print as a direct quote, so you and the speaker can see exactly what was said before deciding how to present it. Use it too when the recording is a record in its own right: a conversation that may matter legally, or research where how something was said is the point. Rev's article makes the same point: printed interviews and legal documents need a full record of everything said. A practical middle path for authors is a clean verbatim transcript with the few passages you might quote marked and redone in full verbatim.

The same passage in each style

These are examples we wrote for this guide. They are not from a real interview. The speaker is a fictional clinic owner talking about her first year.

Full verbatim

Clean verbatim

Edited (non-verbatim)

The full verbatim shows the hesitation and the self-correction from nine to eleven, which is worth knowing if you ever print the number. The clean verbatim keeps "sitting in the car in the parking lot every morning," the detail that makes the story. The edited version is accurate but has already lost it.

A transcript template you can copy

Paste this into a blank document and fill in the header before you start the first pass. The header is what lets you find a transcript six months from now and know exactly which recording and chapter it belongs to.

The examples above are fictional. Keep the brackets exactly as they are so you can search the document for "[" and find every unconfirmed word in one pass.

How to check accuracy without re-listening to everything

You do not need to listen to a two-hour interview twice. Spot-listen to the places where errors cost the most:

• Names. Search the transcript for every name on your names list and for capitalized words that are not on it. A new capitalized word is often a misheard name.
• Numbers. Years, dollar amounts, percentages, team sizes and ages. "Fifteen" and "fifty" sound alike, and a wrong number in a book is the error readers notice.
• Quotes you will print. Listen to every one at normal speed, start to finish, and compare it word for word. If you are quoting someone other than yourself, upgrade that passage to full verbatim.
• Every bracket. Search for "[" and go to each timestamp. Resolve what you can; ask the speaker about the rest.
• Speaker changes. Skim for places where a label looks wrong, such as the interviewer suddenly telling the author's story.
• A random sample. Pick three two-minute stretches at random and check them closely. If they are clean, the rest probably is. If they are not, the whole file needs another listening pass.

Quoting people from a transcript

If you interviewed other people for your book, such as clients, colleagues or experts, their words in your transcript are not yet words you should print. Good practice:

1. Tell them at the start of the interview that you are recording for a book and may quote them, and record their answer.
2. Send them the exact passage you plan to print, in context, before the book goes to the editor. Ask them to confirm it is accurate and that they are happy to be quoted by name.
3. Keep the OK in writing, an email is fine, and note it in the transcript header.
4. Tidy lightly, and only with agreement. Removing ums from a quote is normal. Changing what someone said is not. If the clean version changes the meaning in any way, ask.
5. Offer anonymity when a story is sensitive: change the name and identifying details, and say so in the book.

This is good practice, not legal advice. If a quote is sensitive, critical of someone, or from a person who may object, talk to a lawyer before you print it.

Mistakes to avoid

• Editing during the correction pass. Fixing grammar while you listen blurs what was said with what you wish was said. Correct first, edit later.
• Losing the link to the audio. A transcript with no timestamps and a vague file name is very hard to check later.
• Transcribing everything at the same level. A ten-minute tangent can be one line: [tangent about hiring, 00:14:20 to 00:24:10].

If you would rather not do any of this yourself, an interview-based ghostwriter handles the recording, transcription and writing. You can book a free consult with Scribe Media to talk through whether that fits your book.