Best transcription apps for research interviews in 2026

Short answer first: for thesis interviews, pick a tool that labels speakers, exports to Word and keeps the audio somewhere your ethics approval allows. MacWhisper runs on your own Mac, free or 64 euro once for Pro. Good Tape processes in the EU and costs 16 euro a month on the yearly plan for 20 hours. If your university licenses Microsoft 365, Word's Transcribe covers 300 minutes a month at no extra cost. Otter and TurboScribe are cheap per hour, but read their data pages before a participant's voice goes into either.

The figures below come from each vendor's pricing page, help centre, privacy or security page and support documentation, all read on 25 September 2026. Where a vendor leaves a number out, the table marks it not published.

What does an interview transcription tool need to handle?

A research interview is not a lecture. There are two or more voices, the content is personal, and the transcript becomes data you will quote in a thesis. The tool that suits a lecture can be the wrong one here. For a wider comparison aimed at lectures and group work, the guide to transcription apps for students covers the general case; this one stays with interviews.

Where does the audio go, and does the participant know?

Your participant agreed to be recorded for a stated purpose. If the recording then goes to a server abroad, or into a vendor's model training, the information sheet should have said so. Read the vendor's privacy or security page for three things: the country or region of storage, whether content trains its models, and how long files are kept.

Does it separate the speakers?

An interview transcript without turns is hard to code. Speaker labels mark where the interviewer stops and the participant starts, and a tool that lets you rename "Speaker 1" to a pseudonym saves an afternoon per interview. Not every tool offers this, and some keep it for paid plans.

Does it handle your language and your existing files?

Many theses involve interviews in the participant's first language. Check the language list against the languages you actually interview in. Also check uploads, since interviews are often recorded on a phone or a dictaphone and transcribed later, and some tools cap either the length of each file or the number of files a month.

What does an hour of audio cost, and can you export to Word?

A study with twelve one-hour interviews needs twelve hours of transcription, often inside one or two months. Divide the monthly price by the hours included to compare tools fairly. Then check the export: a .docx file is the usual route into coding software and into an appendix.

How do the tools compare for interviews?

The table sets the seven tools against the criteria above. Where the tool sections give a price per hour, it assumes you use the whole monthly allowance.

Notibo is the only tool here that does not list speaker labels as a feature of a plan, and MacWhisper is the only one that keeps the audio off a vendor's server. Those two facts sort most of the table before price comes into it.

What do Otter and Notta offer for interviews?

Otter and Notta are both meeting tools first. They do handle uploaded interviews, and both label speakers, but their free plans are built for short calls, and neither stores data in Europe.

• Otter's free plan stops at 30 minutes per conversation. The Otter pricing page gives Basic 300 minutes a month but only 3 file imports in total, and Pro 1,200 minutes, 10 imports a month and 90 minutes per conversation.
• Otter Pro is 16.99 US dollars a month, or 8.33 billed yearly. The same page lists six transcription languages and puts .docx export on Pro and above.
• Otter trains its own models on de-identified data. Its privacy and security page says it uses AWS storage, de-identifies user data before training its models, and reminds users to ask for consent before recording others.
• Notta's free plan transcribes 3 minutes per recording. The Notta pricing page gives Free 120 minutes a month in 3-minute pieces, and Pro 1,800 minutes a month, 100 uploads and up to 5 hours per recording.
• Notta Pro is 13.49 US dollars a month or 97.99 US dollars a year. Notta's help centre article on plans and pricing prints the monthly price, and the pricing page shows the yearly total as 8.17 a month billed annually. Its language article puts monolingual transcription at 58 languages.
• Notta stores data in Tokyo. Its help centre article on data security says Notta uses AWS and that your data sits in a data centre in Tokyo.

For interviews, Otter's 90-minute cap on Pro is enough, and the price per hour is low: 16.99 US dollars for 20 hours is about 85 cents an hour, or about 42 cents on yearly billing. Notta's 1,800 minutes a month come to about 45 cents an hour at 13.49 US dollars a month, or about 27 cents on the yearly plan. The model training on de-identified data is the point to raise with your supervisor. Notta's free plan is effectively a demo at 3 minutes, and its Tokyo storage matters if your approval requires data to stay in a given region.

Is Good Tape the safer choice for European data?

Good Tape is made in Copenhagen and built for journalists and other professionals, which shows in what it puts on its pricing page: data processing, certification and training policy come before features.

• Pro is 20 euro a month, or 16 euro a month billed yearly. Good Tape's pricing page shows 192 euro for the year, includes 20 hours of transcription a month, bills in euro whatever estimate it shows in your own currency, and offers a free trial of 1 hour with no card.
• Processing is in the EU and nothing trains a model. The same page lists EU data processing, ISO 27001 certification, a data processing agreement and no model training on your data.
• Speaker labels and Word export are standard. Pro includes speaker labels, 100+ languages, unlimited file length, and transcript export as .docx, .srt, .txt, .pdf and .tex.

At full use, 20 hours cost 1 euro an hour on monthly billing and 80 cents on the yearly plan, and there is no cap per file, which suits a two-hour interview. The data processing agreement is the document an ethics office may ask for. The limit is volume: a study with more than 20 hours of audio in one month needs either a second month or a larger plan.

Can MacWhisper keep interviews on your own Mac?

MacWhisper is a Mac app that runs Whisper speech models on your own computer. For interviews with sensitive content, that removes the question of where the audio goes, because it goes nowhere.

• The free version transcribes locally. The MacWhisper site lists a free version that records or opens mp3, wav, m4a, ogg, opus, mov and mp4 files, supports 100 languages and exports to .txt and subtitles.
• Pro is 64 euro once. The Pro licence includes lifetime updates and adds automatic speaker recognition, batch transcription and export to .docx, .pdf and other formats. The Gumroad page that handles the sale listed a single Pro licence at 65 euro on the same day, so expect either figure at checkout.
• Cloud AI is optional. The site describes local models for private files, and separate connections to AI providers with your own keys, which you can leave unused.

For a student on a Mac, this is the cleanest answer to an ethics question: the site says local processing keeps the data on your Mac, and a cloud model is only used if you choose one. The costs are a Mac, which Windows users do not have, and processing time on your own machine. Speed depends on the model you pick and the Mac you run it on, and the site publishes no length cap.

What about Word's Transcribe and TurboScribe?

These two sit at opposite ends. Word's Transcribe may already be paid for by your university. TurboScribe is a cheap, fast upload service that keeps its files in the United States.

• Word gives Microsoft 365 users 300 minutes a month. Microsoft's support page on transcribing recordings accepts .wav, .mp4, .m4a and .mp3, lists more than 80 locales, and labels speakers as Speaker 1, Speaker 2 that you can rename.
• Word saves the recordings to OneDrive. The same page says uploaded audio lands in a Transcribed Files folder, that Word for the web needs Edge or Chrome, and that the Windows desktop version of the feature is currently limited to commercial tenants, so check that it appears in your university account.
• TurboScribe Unlimited is 20 US dollars a month. The TurboScribe home page also shows 10 US dollars a month billed yearly, files of up to 10 hours or 5 GB, 98+ languages, speaker recognition and .docx export.
• TurboScribe's free tier is 3 files a day. Each free file can run up to 30 minutes, which fits a short interview but not an hour-long one.

If your university account includes Transcribe, 300 minutes covers five one-hour interviews a month inside the tenant your institution already manages, which many ethics offices prefer. TurboScribe says files are encrypted and only you can access them. Its security FAQ puts its primary data infrastructure in the United States, and its privacy policy says data from users in the EU is transferred to and stored there, so check that against your ethics approval before personal data goes in.

What do your ethics approval and consent form need to say?

Transcription is part of data handling, so it belongs in the same paperwork as the recording itself. Nothing in this section replaces your own university's guidance; it lists what that guidance usually asks you to decide.

• Consent comes first, in writing. Record only people who have agreed to it, and keep the signed consent with the study records. Never record someone who has not agreed.
• The information sheet names the tool. Participants should know the recording will be transcribed, by which service, where the audio and transcript are stored, and when they will be deleted.
• Your university's rules on tools decide. Many institutions keep a list of approved tools for personal data or require a data processing agreement. Check that list, or ask your research data office, before you upload anything.
• Anonymise where the study needs it. Replace names, places and employers with pseudonyms in the transcript, and if the approval does not allow identifiable audio to leave your device, use a local tool for the transcription step.

The practical order is simple: choose the tool, then write the information sheet, then apply for approval, then record. Changing the tool after approval usually means an amendment.

How do you get from a transcript to coding?

Coding starts with a clean document, not with the raw transcript. Whatever tool you use, plan a correction pass per interview before the file goes into analysis software.

Export the transcript as a Word file. Listen through once at normal or slightly faster speed, correct names and technical terms, and fix any turns the speaker labels got wrong. Replace real names with the pseudonyms from your study key, and remove details that could identify the participant. Then save a final version with a consistent file name.

That Word file is the usual starting point for the analysis packages most departments license, such as NVivo, MAXQDA or ATLAS.ti. Each has its own import guide and its own way of reading speaker turns, so check the one your department uses before you transcribe the whole study. The article on transcripts versus notes explains why a verbatim record, not a summary, is what coding needs.

Which app should you use for interview transcripts?

Decide on the data question first and the price second. Most students find that the first question leaves two or three tools, and the second picks one.

If your ethics approval says identifiable audio must not leave your device, use MacWhisper on a Mac, and buy Pro for 64 euro if you need speaker recognition and .docx export.

If your university wants EU processing and a data processing agreement, Good Tape at 16 euro a month on the yearly plan fits, with speaker labels and no cap per file.

If your university licenses Microsoft 365 and approves it for research data, start with Word's Transcribe and its 300 minutes a month.

If your interviews are short, in English and not sensitive, Otter's or TurboScribe's free tiers may cover a pilot interview, but check the data pages against your approval first.

If you want notes from a supervision meeting or a lecture rather than a coded transcript, a note tool fits better than any of these.

Before you commit to a tool for a whole study, transcribe one pilot interview, with the participant's consent, and check the speaker turns and the export against what your analysis needs.

Where Notibo fits

Notibo is not the first choice for coded interview transcripts, and it is better to say so plainly: the transcript is split by speaker only where Notibo can tell the voices apart, and that is not something to build an analysis on. If the turns between interviewer and participant matter for your analysis, one of the tools above will save you work.

Where it can help is the rest of a research project: supervision meetings, seminars, lectures on method, and the reading. You upload an audio file up to 50 MB, or record on a computer in Chrome or Edge, and get a transcript plus a summary, main points and action items, in any of 89 languages. The transcript and notes can be corrected and exported as PDF or Word. The transcription page shows what comes back.

Files are stored in the EU, in Frankfurt, and transcription runs on US processors under standard contractual clauses, as the privacy policy sets out with the processors named. Audio is deleted once the transcript and notes are saved. Pro is free for 14 days, no card needed, and the trial covers 240 minutes of audio. After that the free plan is 30 minutes of audio a month, and Pro is 9.99 euro a month or 88.99 euro a year, with the details on the pricing page. Check your approval before any participant's recording goes into it.