The honest answer: it depends on what's being said
"The average recording" doesn't exist in the abstract. A voice memo, a two-hour lecture, and a full-day deposition are all recordings, and averaging them together produces a number that describes nothing. The useful question is narrower: how long is the average recording of a particular kind – a podcast episode, a meeting, an interview?
Sort the published data that way and a pattern appears. Most spoken-word recordings people actually sit down to transcribe cluster in a 30-to-60-minute band. Around the edges of that band sits a long tail: the marathon research interview, the two-hour class, the deposition that runs a full workday. That tail is small, but it's exactly where length starts to matter.
Every figure below comes from a public, citable source, sorted from shortest typical genre to longest. Last verified: 2026-07-13.
Podcasts: a median just under 39 minutes
Pacific Content analyzed the lengths of more than 10 million podcast episodes and found a mean of 43 minutes 24 seconds and a median of 38 minutes 42 seconds. The median is the figure to trust: a handful of very long shows drag the mean upward, and Pacific Content itself flags the median as more reliable. Call it a little under 39 minutes for a typical episode.
The spread matters as much as the middle. Buzzsprout's stats, drawn from the shows it hosts, put the largest single share of episodes – 31% – in the 20-to-40-minute range. Another 20% run under 10 minutes, 16% land between 10 and 20 minutes, 20% run 40 to 60 minutes, and 13% cross the one-hour mark. So most episodes are short-to-medium, but roughly one in eight is a long-form recording that a length-capped tool can choke on.
If a podcast is what you're working with, the podcast transcript tool and the guide to transcribing a podcast cover the workflow end to end.
Meetings: 51.9 minutes, and a lot of them
Reclaim.ai's Smart Meetings report, based on a survey of more than 1,300 professionals, puts the average business meeting at 51.9 minutes – up 2.6% from 50.6 minutes in 2021. That's survey data, self-reported rather than measured from calendars, so treat it as a sense of the middle rather than a precise constant.
The volume is the striking part. The same report found professionals spend an average of 14.8 hours a week in meetings – 37% of their working time – though that's down from 21.5 hours a week in 2021. At roughly 52 minutes each, that's a lot of hour-ish recordings piling up for anyone who captures their calls.
That volume is why a meeting summarizer exists, and why transcribing a Zoom meeting is one of the most common transcription tasks there is.
Interviews: about an hour, by design
For qualitative research, a review of 227 studies (Young et al., 2018), cited in the International Journal of Qualitative Methods, found the average in-depth interview lasted 64 minutes. That's the longest of the everyday genres, and it fits: an in-depth interview is meant to run until the subject has said everything worth saying.
It doesn't have to. The same IJQM paper (Zavattaro & Gille) argues that a well-designed 30-minute interview can yield data just as strong as a longer one. But if you're planning storage, or a per-file length limit, an hour is the honest number to build around – and a batch of them adds up fast.
Lectures: 50 minutes, unless it's a lab
The standard U.S. college class period, per College Board, runs 50 minutes meeting three times a week, 75 minutes twice a week, or 3 hours once a week. So a single lecture recording is usually somewhere between 50 and 75 minutes – comfortably inside the 30-to-60-minute band, if a little above it.
Labs and discussion sections are the exception. College Board notes these typically run 2 to 4 hours, and a once-a-week lecture block can hit 3 hours on its own. A student recording a full lab is holding one of the longer single files most people will ever transcribe.
Sermons: a median of 37 minutes, split sharply by tradition
Pew Research Center's Digital Pulpit study analyzed 49,719 sermons from 6,431 churches and found a median length of 37 minutes. But the overall figure hides a wide split by tradition: Catholic homilies ran a median of 14 minutes, mainline Protestant sermons 25, evangelical Protestant 39, and historically black Protestant sermons 54 minutes – nearly four times the Catholic figure.
One caveat worth keeping: this is the length among churches that publish their sermons online, which skew toward larger, urban congregations. It's a strong read on recorded sermons, not a census of every pulpit in the country.
Sales calls: length doesn't predict the outcome
Sometimes the interesting number is the one that isn't there. Gong analyzed 30,000 first sales calls and found no statistically significant correlation between how long a call ran and whether it earned a second meeting. A longer discovery call is not a better one.
Gong found the opposite of what you'd expect on scheduling, too: offering a 30-minute slot made prospects 12% more likely to actually show up than offering a 60-minute slot. Shorter, in this genre, tends to win – which keeps most sales recordings on the shorter end of the band.
The long tail is where length actually bites: depositions run 7 hours
At the far end of the distribution sits the legal deposition. Under the U.S. Federal Rules of Civil Procedure, Rule 30(d)(1), a deposition is limited to one day of 7 hours unless the court orders or the parties agree otherwise. That's a single recording the length of a full workday.
This is the whole point of the distribution. The median case – a 40-minute podcast, a 52-minute meeting – is trivial for any tool. It's the tail that separates them: the two-hour lecture, the 64-minute interview times forty subjects, the seven-hour deposition. A per-file minute ceiling or an upload-size limit never bites on the median. It bites here, on the recordings that matter most and that you can least afford to lose halfway through.
What this means if you have to transcribe them
If your recordings sit near the median, almost anything works, and length is a non-issue. The reason to think about it at all is the tail: the moment you're holding a two-hour file or a stack of hour-long interviews, per-file caps and monthly minute limits stop being fine print and start being the thing that decides whether the job gets done.
That's the case Pepys is built for. Credits are pay-once and never expire, there's no per-file length cap, and a long recording is split into chunks, transcribed, and stitched back together with the timestamps intact – so a seven-hour deposition is a bigger file, not a wall. If you want the mechanics, see how to transcribe long audio and, for the time side of it, how long it takes to transcribe an hour of audio. Pricing is on the pricing page.