Transcript formatting: speaker labels, timestamps and house style

Verbatim and clean-read

This is the first choice, and the one that changes the finished document most.

Verbatim keeps what was said. False starts, repetitions, filler words, stammers and interjections all remain. The transcript records the speech rather than tidying it.

Clean-read removes them. False starts are dropped, repetitions collapsed, filler taken out. The words themselves are never changed, but the document can be read at speed.

The right choice depends on what the transcript is for. Where how something was said matters, verbatim is the only defensible option: legal proceedings, investigative interviews, research where hesitation and self-correction are themselves data. Where the transcript will be circulated and read as a document, clean-read serves the reader better.

If you do not specify, we will ask. It is not a choice we make on your behalf.

Speaker identification

Named labels use the person's name. Role labels use their function, such as Interviewer and Respondent, or Chair and Member.

Names are clearer where readers know the participants and the record needs to attribute statements to individuals. Roles are better where the transcript will be shared beyond the room, where participants were promised a degree of anonymity, or where the function matters more than the person.

Where a speaker cannot be identified with certainty, they are labelled consistently as an unidentified speaker and kept distinct from the others, rather than guessed at. A wrong attribution is worse than an unnamed one, because nothing in the document shows it is wrong.

Introductions at the start of a recording make all of this substantially more reliable, which is why we ask for them.

Timestamps

Interval stamps appear at a fixed cadence, commonly every thirty seconds or every minute. They suit long recordings and make it straightforward to check a passage against the audio.

Speaker-change stamps appear each time a new person begins. They suit interviews and panels, where the structure of the conversation is what you are navigating.

The question to ask is whether anyone will need to return to the audio. If they will, timestamps are not decorative. If the transcript is the final artefact and the recording will not be revisited, they add clutter.

Tell us the interval you want at the outset. Changing it afterwards means re-checking the whole document.

Marking what cannot be heard

Where a passage genuinely cannot be made out, it is tagged as inaudible with a timecode. Where speakers talk over each other, it is tagged as crosstalk, also with a timecode.

The timecode is the point. A tag without one records that something was lost. A tag with one lets someone go to that moment and try to resolve it, often successfully, because they were in the room and we were not.

This is the same principle that runs through everything else here. A marked gap is honest and recoverable. A confident guess is neither, because nothing in the document tells a reader to check it.

A transcript carrying a handful of properly placed inaudible tags is a better document than one carrying none, if the audio was difficult. The absence of any tags on a poor recording tells you something about how it was reviewed.

Sector conventions

Legal work uses a question-and-answer structure for examinations, with questions and answers labelled rather than named, and exhibit references reproduced exactly as stated on the record.

Medical work follows terminology conventions for drug names, dosages and anatomical terms, where an approximate spelling is not acceptable and the term is verified rather than approximated.

Academic and research work often requires non-verbal utterances, pauses and overlaps to be marked, because they are part of what is being analysed rather than noise to be cleaned away.

Media work is checked against reading-speed and line-length limits where captions are produced, so the files are usable on delivery rather than after your editor has fixed them.

These are defaults. Where your field or your organisation does it differently, your way takes precedence.

Telling us your house style

If you have a style guide, send it. We will work to it.

If you have a glossary of recurring names, products or terms, send that too. It is the single most useful document a client can provide.

If you have neither, we apply our own conventions, and apply them consistently across every file in the job. Consistency is what makes a set of transcripts usable together.

Style decisions are inexpensive before work starts and expensive afterwards. Raise them at quote time.