Free tools
How long does transcription take?
Enter the length of the recording and how it was recorded. You get estimates for typing it out manually against automatic transcription with editing, and what the difference is worth.
The estimates are based on experience, not guarantees. The single factor that moves them most is how close the microphone was to whoever was speaking.
What governs the time it takes
| Material | Manual per hour of audio | Automatic, your editing |
|---|---|---|
| One speaker, good audio, script | 1.5 to 2 hours | 10 to 15 min |
| Interview, two people, good audio | 2 to 3 hours | 20 to 30 min |
| Meeting, several speakers, ordinary audio | 3 to 4 hours | 30 to 45 min |
| Poor audio, dialect, overlap | 5 hours and up | 45 to 90 min |
Why manual transcription takes so long
Because you hear faster than you type. Ordinary speaking pace is 120 to 150 words a minute. Ordinary typing speed is 40 to 60. The difference is covered by pausing and rewinding, and every pause costs two to three seconds on top of the typing itself.
It is also why the simplest time-saving move is to set the playback speed down to 0.75x, not up. You save more by not having to rewind than you lose on the slower pace.
Where the automatic time actually goes
- Proper nouns and place names. The single largest item. Surnames are nearly impossible to tell apart by ear.
- Jargon and abbreviations. Internal language and product names are the weakest flank in every system.
- Telling speakers apart on a shared microphone. With five people around one phone, the boundaries between them blur.
- Overlapping speech. When two people talk at once, one of them usually disappears.
Ordinary words are usually right. The editing is therefore targeted reading, not rewriting.
How to halve the editing time
- Put the microphone near whoever is speaking. Everything else comes second.
- Say the names out loud at the start of the recording. It makes telling speakers apart much easier.
- Ask people to speak one at a time. Overlap is where every system fails.
- Turn off fans and ventilation. Steady noise is worse than short sounds.
- Choose a tool that is actually trained on your language, not one that merely lists it.
About transcription time
Manually it takes four to six hours for someone unpractised and two to three for someone practised. Automatically it takes a few minutes of machine time, plus 20 to 40 minutes of editing per hour of audio depending on audio quality and the number of speakers.
Audio quality most, then the number of speakers and how specialised the jargon is. One person speaking clearly into a good microphone is fast. Five people around one microphone in a room with hard acoustics is many times the work.
Five to ten times faster measured in your own working time, depending on the material, because you edit instead of type. The machine time itself is a few minutes per hour of audio and requires no attention from you.
Because the errors cluster in proper nouns, place names and jargon, which is precisely where the content lives. Ordinary words are nearly always right. The editing is therefore targeted reading, not rewriting.
No. The calculator works in your browser and uploads nothing. It does not accept audio files either, only the length of the recording.
Speak. It is written.
Sign in with email: no passwords, no credit card. 14-day free trial.