How long does it take to transcribe 1 hour of audio?

If you’ve got an hour-long interview, lecture or meeting to turn into text, the answer depends mostly on one question: are you typing it yourself, or letting software do it? By hand, an hour of audio usually takes several hours. With automatic transcription, it takes minutes, plus however long you spend checking the result.

Here are the numbers, where they come from, and what makes them go up or down.

The short answer

Method	Typical time for 1 hour of audio
Professional transcriptionist	About 2 to 4 hours
Non-professional typing it yourself	About 4 to 8 hours, more with poor audio
Automatic (software) transcription	Minutes, plus time to review

These ranges come from transcription companies that do this work for a living. Rev says the average time for a professional transcriptionist is 2 to 3 hours per audio hour, that the average person takes about 4 hours, and that poor audio quality can double or triple the time. GoTranscript puts skilled transcriptionists at about three to four hours on average and beginners at six to eight hours, and notes that automated systems may finish in minutes. These are industry estimates rather than controlled studies, but they agree closely.

Why manual transcription takes so long

The basic problem is that people speak much faster than they type.

Speaking speed vs typing speed

Conversational speech is often quoted at around 150 words per minute, though it varies a lot from person to person and between a relaxed chat and a rushed presentation. You can measure a real recording with our speaking rate calculator. At around 150 words a minute, an hour of continuous talking is roughly 9,000 words.

Typing is slower. A large study from Aalto University, which analysed the typing of volunteers in an online test, found that average users typed 52 words per minute, with most people between 30 and 60.

Put those together and typing 9,000 words at 52 words a minute takes almost three hours, if you could type continuously without ever stopping. You can’t. Transcription means listening, pausing, typing, rewinding to catch a word you missed, and listening again. That stopping and starting is why even professionals, who type faster than average and use foot pedals and keyboard shortcuts, still need two to four hours.

Things that make it take longer

Rev and GoTranscript both point to factors like these:

• Audio quality. Background noise, echo, a distant microphone or a bad connection means more rewinding. This is the biggest single factor.
• Number of speakers. More people means more crosstalk, and you have to label who said what.
• Accents and fast speech. Unfamiliar accents and people who talk quickly need more replays.
• Specialised vocabulary. Technical terms, product names and people’s names have to be looked up or checked.
• Style of transcript. A strict verbatim transcript (every “um”, false start and stutter) takes longer than a clean one. Adding timestamps and speaker labels adds time as well.

How long automatic transcription takes

Speech recognition software turns audio into text much faster than a person can. For a one-hour recording, the processing itself generally takes minutes rather than hours. Exactly how many minutes depends on:

• Where it runs. Cloud services run on powerful servers but you have to upload the file first, which on a slow connection can take longer than the transcription. Apps that transcribe on your computer skip the upload but use your processor, so a faster computer finishes sooner.
• The model. Larger, more accurate models are usually slower than small, fast ones.
• Extras. Working out who spoke (speaker diarization), adding punctuation and timing every word all add some processing.

Some tools also transcribe live, while the meeting or lecture is happening, so the text is ready when the recording ends. Hark does both: it shows a live transcript during a meeting, and then makes a more careful full transcript after it ends.

Don’t forget review time

Automatic transcription is fast, but it isn’t perfect. Expect to spend some time checking it, particularly for:

• names of people, companies and products;
• jargon and acronyms;
• numbers, dates and amounts;
• places where people talked over each other.

How long review takes depends on how good the audio was and how accurate you need the transcript to be. For your own meeting notes, skimming for obvious errors may be enough. For a quote you’re going to publish, check it against the audio. Tools that let you click a word to jump to that moment in the recording make this much quicker.

Choosing the right approach

Type it yourself if the recording is short, the audio is difficult, or you have no budget and plenty of time. Use a media player with keyboard shortcuts for pause and rewind, and slow the playback down.

Pay a human service if you need high accuracy for publication, legal or research use, and the cost is justified. Even then, turnaround is typically measured in hours or days, not minutes.

Use automatic transcription for most everyday needs: meetings, interviews, lectures and voice memos. It’s dramatically faster, and fixing a few mistakes is far quicker than typing everything.

Making automatic transcripts better

A few habits improve accuracy with any software:

1. Use a decent microphone, close to the speaker. A headset or desk mic beats a laptop mic across the room.
2. Record each side separately if you can. Separate tracks for your microphone and the other people on a call reduce crosstalk.
3. Reduce background noise. Close windows, mute when you’re not speaking, and avoid echoey rooms.
4. Ask people not to talk over each other. Overlapping speech is hard for software and people alike.
5. Keep a list of names and terms. Some tools let you give them a vocabulary list to get names and jargon right.

Where Hark fits

Hark is our desktop app for Windows, macOS and Linux. It transcribes on your computer, so there is no upload and no per-minute cost, and it labels who spoke by their voice. You can record meetings directly or import existing recordings (mp3, m4a, wav, mp4 and more). It is English only, and the speech models are a one-time download of about 730 MB. Transcription is free and unlimited.

If you have a specific file to convert, our guides cover transcribing an audio recording to text, mp3 to text and transcribing an interview.

Summary

• By hand: roughly 2 to 4 hours per audio hour for a professional, and 4 to 8 hours or more for everyone else.
• Why: people speak around three times faster than the average person types, and transcription means constant pausing and rewinding.
• Automatically: minutes of processing, plus review time that depends on audio quality and how accurate you need the result to be.

Whichever way you do it, if the recording includes other people, make sure they agreed to being recorded. Some places require everyone’s consent.