Skip to main content
How to Write a Transcript: Format, Rules and an Example
2026/11/17

How to Write a Transcript: Format, Rules and an Example

How to write a transcript: pick verbatim or clean style, set up speaker labels and timestamps, mark unclear audio and proofread. With a copyable template.

To write a transcript, choose a style first (verbatim or clean verbatim), then write down the speech turn by turn, starting each turn with a speaker label and adding timestamps at speaker changes or at regular intervals. Mark unclear words as [inaudible] with a time, and finish by proofreading against the audio. The fastest way is to edit an automatic draft instead of typing from scratch.

The rest of this guide explains each decision, gives a copyable template and an example, and lists the mistakes that make a transcript hard to use.

Step 1: Decide what the transcript is for

The purpose decides how detailed the transcript must be. Answer these questions before you type anything:

QuestionIf yesIf no
Does every hesitation matter (legal, linguistics, some research)?Use verbatimUse clean verbatim
Will someone need to find quotes in the audio again?Add timestampsTimestamps are optional
Are there two or more speakers?Add speaker labelsA single block of text is fine
Will it be published or shared outside your team?Get consent and check namesInternal notes need less polish

Step 2: Choose a transcription style

There are three common styles. Pick one and use it throughout.

  • Verbatim (true verbatim). Every word and sound as spoken: "um", "uh", "you know", false starts, repetitions and stutters. Use it for depositions, police interviews, conversation analysis and some qualitative research.
  • Clean verbatim. The speaker's words and meaning, without fillers and false starts. Grammar stays as spoken. This is the default for interviews, podcasts, meetings and lectures.
  • Edited (intelligent) transcript. Lightly rewritten for reading: broken sentences are fixed and repetition is removed. Use it for articles and summaries, and say that it has been edited.

The same sentence in each style:

StyleText
Verbatim"So, um, we—we basically, uh, we moved the launch to, like, March."
Clean verbatim"So we basically moved the launch to March."
Edited"We moved the launch to March."

Step 3: Set up the transcript layout

A readable transcript has four parts.

Start with the details someone will need later:

  • the title or topic of the recording,
  • the date and place it was recorded,
  • the names and roles of the speakers,
  • the length of the recording, and
  • the style you used (for example, "clean verbatim").

Speaker labels

Start every new speaker turn on a new line with a label. Use full names the first time if you have consent, then a short form. In research, use neutral codes such as Interviewer and P07 to protect identities.

Be consistent. Don't switch between "Maria", "M." and "Speaker 2" for the same person.

Timestamps

Use the format [hh:mm:ss] and put a timestamp at each speaker change, or every 30 to 60 seconds in long monologues. That is enough to find any quote in the audio quickly.

Notation for non-speech

Use square brackets for anything that isn't a spoken word:

NotationUse it for
[inaudible 00:14:02]A word you can't make out, with the time
[Hartley?]Your best guess at an unclear word
[crosstalk]Two people speaking at once
[laughs], [pause]Sounds and silences that change meaning
[phone rings]Interruptions that explain a gap

Keep the list short. Only note sounds that change how the words should be understood.

Step 4: Write the transcript

Work in short passages:

  1. Listen to 5 to 10 seconds, pause, and type what you heard.
  2. Replay before you move on if you're unsure of a word. Slow playback to 0.75× for fast speakers.
  3. Add a timestamp at each speaker change.
  4. Leave a marker such as [check] for names, numbers and terms you need to look up, and keep going. Stopping to research every term slows you down more than a second pass.

Typing a long recording by hand takes many times its running time. If you have more than a few minutes of audio, an automatic first draft saves most of that time. Upload the recording to a tool such as FastScribe's audio to text converter and you get a draft with a time range on every sentence and speaker labels for uploaded files. Then do steps 5 and 6 on the draft instead of typing from zero.

Step 5: Proofread against the audio

Read the whole transcript while listening, ideally at normal speed. Focus on what automatic drafts and tired typists get wrong:

  • names of people, companies and places,
  • numbers, dates and money amounts,
  • technical terms and acronyms,
  • who said what when people interrupt each other, and
  • negatives ("can" versus "can't" changes the meaning completely).

Resolve every [check] marker. Keep [inaudible] only where the audio really is unclear.

Step 6: Save it in the right format

  • DOCX for documents you will edit, comment on or print.
  • TXT for plain text you will search, paste or analyze in other software.
  • SRT if the text will be shown as captions on a video. Each line then keeps its start and end time.
  • XLSX if you want one row per sentence to sort, filter or code.

FastScribe exports all four. See export formats for what each one contains.

Transcript template

Copy this layout and replace the details:

Title: Product team interview: onboarding redesign
Date and place: 12 November 2026, video call
Speakers: Interviewer (Sam Ortiz), P07 (customer, operations lead)
Length: 00:42:18
Style: Clean verbatim. Timestamps at each speaker change.

[00:00:04] Interviewer: Thanks for joining. Can you walk me through the first
day you used the product?

[00:00:11] P07: Sure. The first thing I did was invite my team, and that's where
it got confusing, because the invite email went to spam for two of them.

[00:00:26] Interviewer: What did you do then?

[00:00:29] P07: I sent them the link on Slack [laughs]. Not ideal. After that
the setup was fine, but the import step asked for a CSV format we didn't
have [inaudible 00:00:41].

[00:00:44] Interviewer: Which format were you exporting from?

[00:00:47] P07: From [Hartley?]. I'll double-check the name and send it to you.

Common mistakes to avoid

  • Mixing styles. Removing "um" in one section and keeping it in another makes quotes unreliable.
  • Correcting the speaker. In clean verbatim you remove fillers, not opinions or grammar. Don't improve what someone said.
  • Missing timestamps. Without them, checking a quote means listening to the whole recording again.
  • Guessing silently. If you're not sure of a word, mark it. A wrong name stated with confidence does more damage than a visible question mark.
  • Skipping consent. Ask before you record, and before you publish a transcript that names people.

Summary

A good transcript follows one style, labels every speaker, has timestamps you can search, and marks what you couldn't hear. Decide the purpose first, use the template above, and proofread names and numbers against the audio. For anything longer than a few minutes, start from an automatic draft in FastScribe and spend your time correcting instead of typing.

For interview-specific advice, see Interview Transcription for Journalists. For timestamp formats in each export, see Transcription with Timecode.

Frequently asked questions

What is the correct format for a transcript?

There is no single official format, but most transcripts share the same parts: a header with the recording details, a speaker label at the start of each turn, a timestamp at regular points, and bracketed notes for anything that isn't speech, such as [inaudible] or [laughter]. Pick one style and use it the whole way through.

What is the difference between verbatim and clean verbatim?

A verbatim transcript keeps everything the speaker said, including fillers like um and uh, false starts and repetitions. A clean verbatim transcript removes those but keeps the speaker's words and meaning. Use verbatim for legal, research or linguistic work, and clean verbatim for articles, notes and most business use.

How long does it take to transcribe one hour of audio?

By hand it usually takes several hours, depending on audio quality, the number of speakers and how fast you type. A common shortcut is to create an automatic first draft, then spend your time checking and correcting it against the audio.

How do I mark words I can't hear in a transcript?

Write [inaudible] with a timestamp, for example [inaudible 00:12:41]. If you can make a reasonable guess, put it in brackets with a question mark, such as [Hartley?], so a reader knows to check it.

Should a transcript include timestamps?

Include them whenever someone will need to find a passage in the audio again, for example to verify a quote or cut a clip. A timestamp at every speaker change, or every 30 to 60 seconds, is enough for most transcripts.

Newsletter

Join the community

Subscribe to our newsletter for the latest news and updates