Audio transcription practice
Press Play, listen, and type what you hear. Pause, go back a sentence and slow down whenever you need to.
The recordings
Business
- Voicemail: delivery change: 65 words, level 1 of 3.
- Memo: parking repairs: 58 words, level 1 of 3.
- Letter: quote follow-up: 73 words, level 2 of 3.
- Meeting notes: budget review: 70 words, level 2 of 3.
- Interview excerpt: 71 words, level 3 of 3.
Medical office
- Referral letter: 65 words, level 2 of 3.
- Clinic note: follow-up visit: 67 words, level 3 of 3.
- Front desk: appointment message: 66 words, level 1 of 3.
Legal office
- File note: client call: 67 words, level 2 of 3.
- Letter: notice of meeting: 66 words, level 3 of 3.
General
- Short story: the lighthouse cat: 56 words, level 1 of 3.
- Local news: library opening: 70 words, level 2 of 3.
How the audio works
Your browser reads each recording aloud with its speech voices, one sentence at a time. A voice built into your device is chosen first, so nothing you type or hear is uploaded; voices marked "online" come from your browser's maker. You can choose a voice and a speed. By default playback pauses after each sentence, like a transcriptionist using a foot pedal; turn that off to hear the recording straight through.
Keyboard controls
Keep your hands on the keyboard: Esc plays or pauses, Ctrl + [ goes back a sentence and Ctrl + ] skips forward. Professional transcription software works the same way with a pedal or function keys.
How transcription is scored
Transcription is scored by word accuracy: 100 minus the word error rate, where wrong, missing and extra words each count once. Capitals and punctuation are not marked, because speakers do not say them the same way. Net WPM uses the same edit-distance formula as other edited-text tasks. Write numbers as words, the way they are spoken.
Twelve recordings in three levels
Recordings range from short voicemails (level 1) to letters, clinic notes and interview excerpts (level 3), in business, medical office, legal office and general topics. Medical and legal recordings are office paperwork for practice, not advice.
Questions
Why is there no sound?
Some browsers have no built-in voice, and some need one to be installed in the system settings. If yours cannot read aloud, the page shows each sentence for a few seconds instead, so you can practise typing from memory.
What is a good word accuracy?
General transcription work usually expects 98% or better on clear audio. Medical and legal transcription expect even higher accuracy, with checking afterwards.
Does it work on a phone?
Yes, as long as the phone has a speech voice, but transcription is much faster with a physical keyboard.