Skip to content
Sonorarium
Back to blog

Verbatim vs Clean Transcription: Which One You Need and How to Get It

The differences between verbatim, clean (intelligent) verbatim and edited transcription, with the same sample passage in each style, when to use each and what automatic transcription gives you.

Sonorarium Team 3 min read

Short answer: a verbatim transcript captures everything exactly as said (fillers, repetitions, false starts); a clean transcript removes those speech marks but keeps what each person said; an edited transcript is rewritten to read like prose. For discourse research, verbatim; for most interviews, theses and journalism, clean; for publishing, edited.

The same passage in all three styles

Verbatim:

Um… well, like, we, we started selling online in, in twenty, in March twenty-twenty, you know, when everything shut down. (pause) And it was, phew, it was crazy.

Clean verbatim:

Well, we started selling online in March 2020, when everything shut down. And it was crazy.

Edited:

We started selling online in March 2020, when the pandemic shut everything down, and the experience was overwhelming.

The clean version is still faithful: it doesn't change the meaning or any important words. The edited version is already the interviewer's own writing.

When to use each

Style What it includes Use it for
Verbatim Everything: fillers, repetitions, pauses, laughter, false starts Discourse and conversation analysis, psychology, evidence in legal settings
Clean verbatim What was said, without fillers or hesitations Qualitative interviews, dissertations and theses, journalism, minutes
Edited Rewritten, polished text Articles, interview books, web content

If you'll use it as data in research, check with your supervisor or your methodology first: switching style halfway through means reviewing everything again.

What automatic transcription gives you

Speech recognition models like Whisper tend to produce something very close to clean verbatim: they punctuate like written text and usually drop many fillers ("um", "like", "you know") and repetitions. That's what most people need, and it saves a lot of work.

That means two things:

  • If you need clean verbatim, just review: fix names, numbers and the odd misheard phrase while listening to the audio.
  • If you need strict verbatim, you'll have to add the fillers, repetitions and pauses relevant to your analysis while reviewing. The synced editor helps: click a sentence and that exact clip plays.

Useful conventions for verbatim transcripts

If your analysis requires it, use simple markers and keep them consistent:

  • (pause) or (3 s) for meaningful silences.
  • [laughs], [coughs], [noise] for non-speech sounds.
  • [inaudible 00:12:45] when something can't be understood, with the timestamp.
  • … for interrupted sentences.
  • [name], [company] for anonymization.

Conversation analysis uses more detailed systems (such as Jefferson's, with overlaps and intonation). If that's you, automatic transcription works as a draft, but you'll need to add the notation by hand.

How to get each style, step by step

  1. Upload the recording to the transcription tool and choose the language.
  2. Review the text while listening in the editor.
  3. For verbatim, add fillers, pauses and markers; for clean, fix errors only; for edited, download as Word and rewrite there.
  4. Download as Word (or TXT) and keep the original audio in case someone needs to check a quote.

FAQ

Is clean verbatim acceptable in a thesis? For most thematic and content analyses, yes. If your methodology analyzes how something is said (not just what), you need verbatim.

Can I get the AI to include every filler? Not reliably: models are trained to produce readable text. Adding them during review is the safest option.

What about quoting in the press? Clean verbatim: you can drop fillers, but not change words or meaning. When in doubt, listen to the audio. More tips in interview transcription.