Transcribing research interviews without a third-party processor
Interview data is personal data, and ethics committees ask where it goes. Transcribe answers simply: nowhere. Recordings are transcribed on your own PC, ready for coding in your analysis software.
What ethics and data protection ask
Data management plans and ethics applications usually have to say where recordings are stored, who processes them, and how participants are protected. A cloud transcription service adds a processor to that answer: a contract, a location, perhaps an international transfer, and a line in every consent form.
With offline transcription the answer stays short. The recordings are transcribed on the researcher's own computer, and no one else receives them. Check your institution's rules on storage and encryption; the transcription itself adds no new recipient.
From recording to codable transcript
Import the interviews
Drop in your recordings. With Pro, queue a whole study at once and let it run in the background.
Label interviewer and participant
Speaker detection (Pro) separates the voices; name them, for example I and P1, and every line follows.
Pseudonymise
Replace names, places and employers with pseudonyms everywhere at once, using find and replace.
Check against the audio
Click any line to hear it. Fix mishearings in place; edits save themselves.
Export for analysis
Plain text with or without timestamps, or Word with Pro, for NVivo, ATLAS.ti, MAXQDA or your own coding.
Useful in research
100 languages
Interviews in the participants' own language, transcribed in that language.
A folder per study
Sub-folders for waves, sites or groups; notes beside each interview for memos.
Custom words Pro
Field-specific terms and the names of places and programmes, spelled right.
Published measurements
Speed and error rates with their method, for your methods section. Benchmarks.
A note on verbatim transcription
Whisper tends towards clean verbatim: it often leaves out fillers such as "um" and false starts. If your analysis needs them, such as in conversation analysis, add them while checking against the audio, and say so in your method.
Questions
Is offline transcription GDPR compliant for research?
Compliance depends on your whole process: consent, storage, access and retention. Offline transcription helps because no recording is sent to a processor, so the transcription adds no data processing agreement or international transfer.
Which format should I export for NVivo, ATLAS.ti or MAXQDA?
Word (DOCX, Pro) or plain text both import well. Keep timestamps if you want to return to the audio later.
Can it transcribe focus groups?
Yes. Tell speaker detection how many people took part. Crosstalk is hard for any system, so plan to check busy passages by hand.
Can I cite the accuracy figures?
You can cite our benchmarks page and its data, with the date and the conditions. For your own study, measure the error rate on a sample of your recordings.
Your recordings stay on your PC.
Free to download. Transcribes without an account, without uploads and without the internet.
Coming soon to the Microsoft Store · Windows 10 and 11