How to Transcribe Interviews Privately and Accurately

Interviews often contain information people shared in confidence: sources’ identities, research participants’ experiences, patients’ details, customers’ feedback. Uploading those recordings to an online transcription service means trusting a third party with them. Here’s how to get accurate transcripts while keeping the audio on your own device.

1. Record well

Accuracy starts with the recording. A few minutes of preparation saves hours of correction.

• Put the microphone close to the speakers. A phone on the table between two people beats a laptop across the room. For remote interviews, record each side locally if you can.
• Choose a quiet space. Turn off fans and air conditioning, avoid cafés, and close windows.
• Do a 10-second test and listen back with headphones.
• Ask for consent to record, and explain how the recording and transcript will be stored.

2. Transcribe on your device

1. Open the free transcription tool.
2. Drop in the recording — MP3, M4A (iPhone Voice Memos), WAV, MP4 and most other formats work.
3. Choose Accurate for important interviews, or Balanced for a faster first pass.
4. Choose the language if the recording is short or mixes languages.

The Whisper speech recognition model runs inside your browser. Once it has downloaded, you can even disconnect from the internet. Long recordings are processed in five-minute chunks, so you can watch progress.

3. Edit for accuracy

• Replay doubtful segments. Click ▶ next to any segment to hear it again.
• Fix names and terms with Find and replace.
• Watch for hallucinations. During long silences or music, speech recognizers sometimes repeat phrases or produce text that wasn’t said. Delete those segments.
• Mark unclear audio as [inaudible 12:34] rather than guessing.

4. Add speaker labels

The tool splits the transcript into timestamped segments but doesn’t identify speakers automatically. A quick method:

1. Type “Q:” and “A:” (or initials) at the start of each segment as you review.
2. Use Find and replace if you labelled a speaker wrongly throughout.

5. Export and store safely

• TXT for analysis software and notes.
• TXT with timestamps for quoting and checking against the audio.
• Word (DOCX) for sharing and editing.

Store transcripts as carefully as the recordings. If participants were promised anonymity, consider redacting names before sharing — save the transcript as a PDF and use our Redact PDF tool.

Why local transcription helps with research ethics

Many university ethics boards and data protection rules (such as the GDPR) require researchers to minimise who has access to identifiable data and to know where it’s processed. Cloud transcription can mean data is processed abroad, kept for model training or shared with subcontractors. Processing on your own device keeps the data under your control and makes your data management plan simpler — though you should still follow your institution’s rules for storage and retention.

Summary

Good recording, a strong speech model, careful editing and safe storage give you accurate transcripts without handing confidential conversations to a third party.