How long does hand-transcribing one audio hour take?
Plan on 4 to 7 hours of typing for every hour of audio, according to the University of Bath research-data guide. Peer-reviewed work sets a wider band: Bailey (2008) in Family Practice reports at least 3 hours per audio hour, rising to 10 hours for fine verbatim detail.
Detail level drives that spread. A clean transcript, with filler words removed, sits near the bottom of the range. A fine-grained version that marks pauses, overlaps and nonverbal cues climbs toward Bailey's 10-hour figure. Audio quality, accents and the number of speakers all add time on top.
Published ranges per single audio hour: - 3 hours: fast, low-detail floor (Bailey, 2008) - 4 to 7 hours: typical academic range (University of Bath) - Up to 10 hours: fine verbatim with visual detail (Bailey, 2008)
Working from a single recording rather than a whole project? See our breakdown of how long it takes to transcribe an hour of audio for the per-file view.
How many interviews go into one research study?
A typical qualitative interview study runs a median of 15 to 31 interviews, based on Vasileiou et al. (2018), which analyzed 214 published studies. Per-journal medians landed at 31 interviews (BMJ), 30.5 (Sociology of Health & Illness) and 15 (British Journal of Health Psychology). Assume one hour each (interview length varies by type) and that is 15 to 31 audio hours.
Could you interview fewer people? Some researchers stop once themes repeat. Guest, Bunce & Johnson (2006) found thematic saturation within the first twelve interviews of sixty. Hennink, Kaiser & Marconi (2017) drew a sharper line: code saturation at nine interviews, yet 16 to 24 interviews for full meaning saturation.
Here's the catch. Even a lean, saturation-driven study of 16 to 24 interviews still leaves you 16 to 24 audio hours to transcribe. Cutting interviews trims the interviewing, not the typing that follows. The transcription workload scales with every recorded hour you keep.
See our full breakdown of how many hours of audio a qualitative study contains to map your own project.
The whole-study math: hundreds of person-hours
Do the arithmetic and a full study runs into the hundreds of person-hours. Take a 30-interview study at one audio hour each: 30 audio hours times the 4 to 7 hours of typing per audio hour from the University of Bath equals roughly 120 to 210 person-hours. Treat this as an illustrative calculation, not a single published figure.
How it scales by study size (one audio hour per interview, 4 to 7 hours of typing per audio hour): - 15 interviews: 60 to 105 person-hours - 24 interviews: 96 to 168 person-hours - 31 interviews: 124 to 217 person-hours
Use Bailey's wider band and the totals stretch further. At 3 to 10 hours per audio hour, that same 30-interview study spans 90 to 300 person-hours. The exact number depends on detail level, audio quality and speaker count, but the order of magnitude holds: this is weeks of full-time work.
Splitting long recordings into manageable pieces helps. See how to transcribe long audio for a chunking workflow.
From person-hours to working weeks and days
Convert person-hours to a calendar and a full study eats 3 to 5 working weeks. The U.S. standard workweek is 40 hours (29 CFR 778.101). A 30-interview study at 120 to 210 person-hours therefore fills about 3 to 5.3 weeks, or roughly 15 to 26 eight-hour working days.
Treat these as illustrative conversions, not published totals. The 40-hour week is the federal maximum-nonovertime standard; splitting it into an 8-hour day is a common convention. One person rarely transcribes eight focused hours a day, so real calendars stretch longer once you add breaks, review and correction passes.
One of the biggest variables is you. Outsource to a professional service and the calendar shrinks, but the person-hours do not: someone still types every recorded minute. That labor is why transcription, not interviewing, is often the longest phase of qualitative fieldwork.
How can you cut the transcription time down?
Automatic transcription flips the ratio. Instead of the 4 to 7 hours of manual typing per audio hour reported by the University of Bath, software like Pepys drafts a full transcript in minutes. You then spend your hours checking and correcting, not typing from a blank page. The speech-recognition research is public, including the OpenAI Whisper paper (Radford et al., 2022).
Budget for the human pass anyway. Automated drafts still need correction for accents, crosstalk, jargon and speaker labels, the same things that pushed manual verbatim toward Bailey's 10-hour figure. The win is sequence: you edit a draft instead of building one, which is where the hours disappear.
Two habits protect accuracy. First, follow a consistent method: our guide to qualitative research transcription covers verbatim conventions and speaker labels. Second, plan the next step early, because analyzing interview transcripts is faster when the transcript is clean and searchable from the start.