Transcribing research interviews on your own device
An interview recording holds a participant’s voice and whatever they trusted you with. This guide explains why universities ask where that audio goes, what to ask of any transcription tool, and how to make a draft transcript without sending the recording anywhere.
Last updated: 5 min read
This is not legal or ethics advice. Your ethics board, your institution and the law where you work decide which tools you may use. This guide reports what the linked sources say and what Sessa does.
Why do ethics boards and university IT care where the audio goes?
When you promise a participant confidentiality, the promise covers every copy of the recording. A transcription service that works on its own servers receives a copy. Someone then has to answer for where that copy is kept, who can reach it, how long it stays and whether it is used for anything else.
Universities have begun to answer in writing, and their answers differ:
- Northwestern University’s IT guidance says that some widely used online transcription services are not to be used with its research data.
- The University of Auckland keeps a list of approved ways to transcribe. A speech model installed on a university device is one of the options on it.
- The University of York lists a transcription tool that does all its work on the local device.
- The University of Mannheim’s Research Data Center points researchers to tools that run on their own computer.
Researchers have built tools for the same reason. The paper that introduces aTrain, from the University of Graz, describes a free tool made so that interviews need not be uploaded to be transcribed.
No single rule holds everywhere. The common pattern is that the tool must be one your institution accepts, and that you can say exactly where the recording goes.
What should you ask of any transcription tool?
These questions apply to every tool, Sessa included. Write the answers down: your ethics application and your participant information sheet will probably need them.
- Where is the recording processed: on your device, or on someone else’s server?
- Does anyone other than you receive the recording or the transcript?
- Where are they stored, and are they encrypted?
- Is anything you record used to train or improve a model?
- Can you delete the recording and the transcript, and is the deletion real?
- Does it need an account, and what does the account hold about you?
- Has your institution approved it, or does it need approval before you use it?
A tool that works on your own device gives short answers to the first four. It does not answer the last one. Data-protection law and your institution’s rules still apply to a recording that sits on your own phone or laptop.
How do you transcribe an interview with Sessa?
Sessa is a free app for iPhone (iOS 27 or later) and Mac (macOS 27 or later). Everything is transcribed on your own devices. Your recordings and transcripts stay on the device unless you turn on iCloud sync. If you do, they are encrypted on your device before they are stored in your iCloud, with a key that only your devices hold. Codonic never receives them, and Apple cannot read them.
- Before the interview, install Sessa and download the language pack for the language you will speak. This is the one step that needs a network, and each pack is about 1.1 to 1.6 GB.
- Ask for consent to record, in the way your ethics approval describes.
- Record on the iPhone, or on the Mac. If the interview is a call held on your Mac, Sessa for Mac records your microphone together with the Mac’s sound, so both sides are captured. If you already have a recording, import the file instead.
- Choose the spoken language for the recording and let the device transcribe. It takes minutes, and longer for long recordings: about a quarter of the recording’s length on a recent iPhone.
- Check the draft. Sessa marks the parts it is unsure of. Listen to them, correct the words and name the speakers, for example “Interviewer” and “Participant 7”.
- Export the transcript as a file for your analysis. From then on the file is outside Sessa, and it is yours to protect.
- Delete the recording when your data plan says so.
Sessa needs no account, has no ads and no analytics, and iCloud sync is off unless you turn it on. For the full path of a recording, and ways to check it yourself, see Where your audio goes.
What are the limits?
- The transcript is a draft. It can have wrong words and wrong speakers. Accuracy depends on the room, the microphone and how people speak. Check it against the recording before you quote or code it.
- Seven languages. Arabic, English, French, German, Hebrew, Russian and Spanish. English and Hebrew are the two we have tried the most. The others run on a general multilingual model, so check those drafts with extra care.
- The menus are in English. The app’s interface is in English and Hebrew only, whatever language you transcribe.
- iPhone and Mac only. There is no Android, Windows or iPad version, and older versions of iOS and macOS are not supported.
- No summaries. Sessa writes what was said and who said it. It does not summarise, code or analyse.
- Closed source. You cannot read the code. Some universities prefer open tools for that reason.
- No approval is implied. No university or ethics board has approved Sessa. Ask yours before you use it with participants.
Who is responsible for consent?
Telling the people you record, and getting their consent, is your responsibility. The rules differ by place and profession, and Sessa gives no legal advice.
Transcribing on your own device does not change what you owe your participants. Tell them that you record, how the recording is turned into text, where it is kept and when it will be deleted, as your ethics approval requires.
Sources
- Northwestern University, Information Technology: “AI Transcription of Research Interviews”
- University of Auckland: “Transcription” (research data guidance)
- University of York, IT Services: “Transcribe recordings”
- University of Mannheim, Research Data Center: “Automated speech recognition”
- Haberl and others: “Take the aTrain. Introducing an Interface for the Accessible Transcription of Interviews” (arXiv 2310.11967)
Sources last checked: 16 November 2026