How do you mark inaudible speech in a transcript?

Mark inaudible speech with a bracketed tag, placed where the words were lost, and give it a timestamp. The tag tells the reader that something was said and that the transcriber could not hear it. A guess would look like a fact, so the tag is the honest choice. The exact wording of the tag depends on your style, and the rule that matters most is consistency: one tag set, used the same way in every file.

Use a bracketed tag with a timestamp, at the place of the gap

GoTranscript, a transcription company, advises marking speech it cannot understand as [inaudible mm:ss] with the time ¹. A tag such as [inaudible 01:28] marks a gap at one minute and 28 seconds. The same guide says "Always add the timestamp for easy reference." ² The time is what turns a tag into something a reviewer can act on. Without it, the reader has to scan the whole recording to find the gap.

Put the tag in the sentence where the words should go, not at the end of the paragraph. A reader who sees "We could ship [inaudible 00:23] the second week" knows the missing word sits between "ship" and "the," and can judge whether it changes the meaning.

Count the lost words only when you can count them

Research transcription has a stricter tag. The CHAT manual, the transcription format used by the TalkBank research database, says to "Use the symbol xxx when you cannot hear or understand what the speaker is saying." ³ If you can distinguish the number of unintelligible words, the manual lets you write several xxx strings in a row. ⁴ A single xxx is the safe default, because a count you cannot verify is another kind of guess.

The Minnesota guide uses a different marker from the other two. It says to "leave a space and type [unclear]" when a word or phrase cannot be made out. ⁵ The three conventions do the same job, so choose one and write it into the order. Mixing them in one file is the mistake to avoid. A reader cannot tell whether [unclear] and [inaudible] mean the same thing.

Mark overlapping speech with its own tag

When two people talk at once and you cannot separate their words, use a different tag. GoTranscript advises [crosstalk mm:ss] for that case. ⁶ Crosstalk is not the same as inaudible speech. The first says two people talked at the same time. The second says a word could not be heard. Keep them apart, so a reader knows whether the problem was the room or the recording.

Never guess a word to fill a gap

A guessed word looks exactly like a word the speaker said, so it misleads the reader more than a gap does. The tag is the honest record. If you are not sure whether a word is right, you can write the word and flag it, but a flagged guess still needs a second listen. For a legal or research record, a guess can change the meaning of a quotation, so tag the gap.

It depends: the style decides how much detail the tag needs

• Research and conversation analysis. Use the research convention, with a count where you can count the words. The style guide for the study should name the tag.
• Business and meeting transcripts. A bracketed tag with a timestamp is enough. Readers want to know where the gap is, not how many words it held.
• Legal and court work. The required format sets the rules, and it may not allow the same tags. Ask for the format in the order, and do not invent one. Our guide on verbatim vs clean verbatim transcription explains why the record matters in legal work.
• Captions. Captions are short cues on the screen, so a long tag may not fit. The rules for sound and speech in captions are in the captions and subtitles guide.

What to do when you cannot make out the words

1. Replay the passage with headphones, and slow the playback if your player allows it.
2. Check the speaker's earlier words. A name, a place or a term often recurs, and the context can settle it. If the same word appears in your notes, use the spelling from there.
3. If the words are still unclear, write the tag with the time, using your style's convention.
4. Add the passage to your review list, so a second person listens to it. The question on transcription QA explains how that check works.

Automatic transcription has a related problem. A draft may not mark its own gaps, so listen to the passages where the audio is poor, the speakers overlap or the names are unusual. The question on transcribing low-quality audio covers how to prepare poor recordings before you transcribe them. For the layout of a whole transcript, the question on what a good transcript looks like shows the tags in context. For the term used for the timestamps, see the timestamp glossary entry, and for the wider subject, see our transcription overview.