What transcription means
To transcribe comes from the Latin transcribere: to copy out. In everyday use it means writing out an audio or video recording, so you can reread, search, quote or archive what was said. Think of an interview for an article or thesis, a meeting, a lecture or a podcast.
The word has other meanings too. In linguistics a phonetic transcription records sounds with special symbols, in music it means writing a piece out in notes, and in biology it is the copying of DNA into RNA. This page is about speech to text.
Transcription or dictation?
With dictation you speak text that appears on screen right away, usually short and with one speaker. With transcription the recording already exists and you write it out afterwards, often with several speakers. More about dictation on speech to text.
Types of transcripts
| Type | What it contains | Used for |
|---|---|---|
| Verbatim (full) | Everything: fillers, repetitions, broken-off sentences, pauses, laughter | Conversation and discourse analysis, legal records |
| Clean verbatim | All content literally, without "uh", repetitions and false starts | Most interviews, theses and journalism |
| Summarised | The essence in your own words, per topic | Minutes and meeting reports |
Manual or automatic
Typing out an hour of audio by hand often takes four to six hours: listen for a few seconds, type, rewind, listen again. Automatic transcription with speech recognition does the typing in minutes. What remains is reviewing and correcting.
| Manual (yourself or an agency) | Automatic (software) | |
|---|---|---|
| Time | 4 to 6 hours of typing per hour of audio, or 5 to 10 working days at an agency | At Murmel an hour of audio usually within a minute, then review |
| Cost | Your own time, or about €66 to €120 per audio hour at an agency | €1.45 per hour with prepaid credits at Murmel |
| Quality | An experienced typist hears more in very poor audio | Good with a clear recording, reviewing is still needed |
| Privacy | A person listens to your recording | Software processes the recording, at Murmel in the Netherlands |
How to transcribe a recording
- Make a good recordingOne microphone close to the speakers and little background noise. A good recording saves more correction work than any program.
- Choose the type of transcriptVerbatim, clean or summarised. If it is for research, record your choice in your method.
- Have the recording written outUpload the file or record directly in the Murmel Studio. For a meeting in Teams or Google Meet the note taker can join.
- Name the speakersMurmel separates the speakers. Rename the labels to names or codes, see speaker recognition.
- Review with the audio alongsideClick a word to replay that part. Put recurring names and jargon in the vocabulary so they come out right next time.
- ExportTo Word, PDF, TXT or subtitles (SRT, VTT). See export.
What does transcription cost?
Typing it yourself mostly costs time. A transcription agency charges about €66 to €120 per audio hour. With Murmel an hour of audio costs €1.45 incl. VAT with prepaid credits (from €6.05, valid 6 months), or you take Basic for €10.89 per month incl. VAT with 12 hours. If you transcribe more than about 7.5 hours a month, Basic is cheaper. See pricing.
Transcribing with Murmel
Murmel is built for Dutch speech. The model, murmel-flash, was trained from scratch on Dutch audio, with English as a second language, and delivers a clean verbatim transcript with speakers. Processing happens at Leafcloud in Amsterdam, storage at Scaleway in the Netherlands, and recordings are not used for training unless you choose to allow it. More in security and privacy.
Frequently asked questions
What is the difference between a transcript and a transcription?
In practice none: both words are used for the written-out text of a recording. Transcription can also mean the act of writing it out.
How long does transcription take?
By hand often four to six hours per hour of audio. With Murmel an hour of audio is usually converted within a minute, then you review it.
Is an automatic transcript error-free?
No. With a clear recording it is a good first draft, but names, jargon and people talking over each other need checking. That is why you replay in the editor wherever you are unsure.
Which files can I have transcribed?
Most audio and video files, such as mp3, m4a, wav, mp4, webm, flac and ogg. See audio to text.
May I record someone to transcribe the conversation?
Say beforehand that you are recording and why, and ask for consent when personal data is discussed. For research the rules of your institution often apply too.