How to transcribe audio faster and more accurately

Transcription combines three skills: listening, remembering a phrase long enough to type it, and typing. Most people who struggle are not slow typists; they stop and start the recording too often, or let the recording run on while they fall behind.

Work in phrases, not words

Play a phrase of five to ten words, pause, type it, and play the next. Pausing after every word breaks your rhythm; letting a whole paragraph play means you will forget parts of it. With practice your phrases get longer. TypingSpark's transcription practice pauses after each sentence by default, which is a good starting length.

Keep your hands on the keyboard

Reaching for the mouse to pause is the biggest waste of time in transcription. Learn the playback keys of your software, or use a foot pedal. In our practice tool, Esc plays and pauses and Ctrl + [ goes back a sentence.

Slow down the audio, not your typing

Most transcription software can play audio slower without making it sound deep. Choose a speed at which you can type a phrase in one go, then raise it as you improve.

Type first, check later

If you cannot make out a word, mark it (many style guides use something like [inaudible] with a time) and keep going. Then listen to the whole recording once more while reading your text. Fixing doubts at the end is faster than replaying the same second ten times.

How accuracy is measured

Transcription accuracy is usually the word error rate: wrong, missing and extra words divided by the number of words spoken. 98% accuracy means about one error in every fifty words. Style rules (how to write numbers, whether to keep "um" and "er") vary by client, so check them before you start a real job.

A ten-minute daily routine

• Two minutes of copy typing to warm up, such as a 2-minute test.
• Six minutes of transcription at a speed slightly above comfortable.
• Two minutes reviewing the comparison: which words did you miss, and why?