podcastinterview transcriptfiller wordsqualitative research
In a real interview the person answering does about three quarters of the talking: a median of 72.4% across 1,030 two-speaker recordings. Filler is rarer than people think — about ten “um”s and twenty-four “you know”s in a typical half-hour interview — and a transcript arrives in a median of 5.5 minutes. Those numbers, and what they mean for how you transcribe an interview, are what the third episode of The Transcript Lab is about.
Free
The Transcript Lab is a short podcast from TranscribeNext about what transcription actually does, measured rather than assumed. Every number in the episode comes from data we collected ourselves and publish in full.
Listen and read along
The player below runs the episode next to its transcript, produced by TranscribeNext from this exact MP3. The highlighted word is the one the engine timed at that moment. Click any line to jump to it.
The Transcript Lab · Episode 3
What real interviews sound like · 4:47
0:00 / 4:47
The transcript is what TranscribeNext returned for this exact MP3, with speaker identification on (two speakers) and the speaker labels renamed to the hosts. Two corrections: Chris’s “That’s” at 2:24 and “Everybody’s favorite” at 2:40 had been attributed to Laura. The words are unedited — including 0:18, where “how much um and uh” came back as “how much Um of Israel”: filler words are among the hardest things to transcribe. Numbers are written as digits because that is how the engine wrote them, and word timings are measured, so the highlight follows the words as the engine timed them.
Two speakers, so the episode measures itself: in this transcript Laura holds about 78% of the talking and Chris about 22% — inside the normal range for an interview, where one voice asks and the other answers.
What the episode covers
How long interviews run: across 1,248 two-speaker recordings of five minutes or longer, the median is 29.9 minutes
Talk time: the person who talks most holds a median 72.4% of the talking, 88.7% at the ninetieth percentile. A real 1985 oral history sits at 85.8% to 14.2%
Why the interviewer's questions stay in the transcript word for word, even at about 28% of the talk time
Filler words across 934 English recordings, per thousand words: 2.16 “um”, 0.92 “uh”, 5.20 “you know” — and why the 17.4 for “like” is not a filler count
Verbatim or clean: removing filler cuts about forty words out of four and a half thousand, so choose by your analysis, not to save space
Turnaround: four to six hours per hour of audio by hand; a median of 5.5 minutes automatically across 962 recordings
The real 1985 interview, formatted three ways, and the full data behind every number are in the guide to the interview transcript. For recording habits that make an interview easier to transcribe, see 10 pro tips from 10,000+ interviews.
