Transcribe research interviews on your own Mac

Qualitative research runs on interviews, and interviews have to be transcribed before they can be coded. Doing it by hand takes hours per hour of audio. Transcription services are faster, but most of them upload the recording, and that can clash with what you promised your participants and your ethics committee.

This page covers what to check before you use automatic transcription on research data, and how Hearkeeper, a Mac app that is not released yet, is built to transcribe on your own computer.

Start with what you promised

Before a single interview is recorded, most research involving people goes through some form of ethics review, and participants sign a consent form. Those documents usually say:

• What is recorded, and whether participants may decline recording.
• Who handles the data, including anyone outside the research team, such as a transcription service.
• Where the data is stored and processed, and for how long.
• How it is anonymised before analysis, publication or archiving.

If your consent form says the audio stays within the university, uploading it to a commercial cloud service may break that promise, even if the service is secure. Many universities and funders also ask for a data management plan that sets these things out in advance. The UK Data Service’s research data management guidance is one widely used reference on planning, consent and handling research data.

In some fields the rules are law, not only good practice. In Iceland, for example, health research may not begin until the National Bioethics Committee or another health research ethics committee has approved it (Act no. 44/2014, art. 12). In the EU and EEA, data about health and several other sensitive topics is a special category of personal data under Article 9 of the GDPR.

Follow your own institution’s ethics committee, data protection officer and funder on what you may use. The points on this page are general and are not legal advice.

Transcription that stays on your Mac

Hearkeeper is built so that recording, transcription and notes all happen on your own computer.

Recording in person or online. Hearkeeper records your microphone and your computer’s audio directly on the Mac. It works for an interview in a room and for one over Zoom, Google Meet, Microsoft Teams, Slack huddles, Webex or FaceTime. No bot joins the call.

Transcription on the device. Speech is transcribed on your Mac with whisper.cpp, accelerated by Metal. After you stop, a final pass with Whisper large-v3 turbo produces the transcript and separates the speakers, so interviewer and participant are told apart. On an Apple silicon Mac we measured that pass at about ten times faster than real time.

Nothing uploaded by default. Nothing leaves your Mac unless you connect a team server or your own API key. With neither connected, the audio and transcripts stay on your disk, and it all works offline, including on fieldwork without a reliable connection.

That makes it easier to write a data management plan you can actually keep: the recording is made on a device you control and processed there.

Consent at the start of every recording

Hearkeeper starts every recording with a consent step: read a short notice aloud, post it in the chat of an online interview, or both. A consent receipt is saved with the recording, and a recording pill stays on screen, visible in screen shares, and cannot be hidden.

For research, that step does not replace your signed consent form or your ethics protocol. It is a reminder at the moment it matters, and a record that the participant was told. It also gives the participant a clear moment to say no to recording. The guide on recording meetings with consent covers the legal background.

From transcript to analysis

Check against the audio. Automatic transcription makes mistakes, especially with names, dialects and people talking over each other. Every line of a Hearkeeper transcript links to the second it was said, so you can play the exact moment and correct it before you code.

Search across interviews. Full-text and semantic search run across all your recordings, stored in one SQLite file on your disk. Finding every time a participant mentioned a theme is a search.

Summaries as a starting point. A language model on your Mac can draft a summary of each interview, with no API key. Every line points to its moment in the recording. Treat these as orientation, not analysis: the coding is still yours.

Export to Markdown. Transcripts and notes export to Markdown, plain text that you can open in any editor, anonymise, and bring into the qualitative analysis software your project uses. Nothing is kept in a proprietary format.

Multilingual interviews

Transcription covers 99 languages. Icelandic is first-class, and interviews that switch between Icelandic and English are detected automatically, with each part kept in its own language. That helps cross-national studies and interviews with participants who move between languages. See Icelandic transcription for more.

Your responsibilities do not go away

Local processing narrows who can reach the data. It does not do the rest of the work for you:

• Secure the device. Use disk encryption and a strong password on the Mac that holds the data.
• Plan backups in line with your data management plan.
• Anonymise transcripts before sharing or archiving them.
• Delete recordings when your protocol says to. If you never connected a server, deleting on your Mac leaves no cloud copy.

Hearkeeper has no compliance certification of any kind. Whether it fits your project is for you and your institution to decide.

When is it available?

Hearkeeper is coming to the Mac first (macOS 14.2 or later, Apple silicon and Intel), with other platforms later. It is made in Reykjavík by Drangalia ehf.

If you want interview transcripts that stay on your own computer, join the waitlist and we will email you when there is a build for you.