Guide
The best transcription software for qualitative research interviews
For a one-off study, transcribe on your own machine with noScribe for free or MacWhisper for €65 once. For a repeat programme where interviews accumulate and get re-quoted, Phonotheca keeps every transcript in one searchable archive for $12 a month. Where every word must be right the first time, Rev's human service charges $1.99 a minute.
By Stefano, who builds Phonotheca. Prices shown as of August 2026.
Four things decide the choice
Speaker separation. A two-person interview transcribed as one block of text costs more correction time than any number of misheard words. Test a tool on your hardest recording, not your cleanest one.
Hours or files. Some plans cap uploads per month rather than audio duration. A study of thirty short interviews hits a file cap that a study of five long ones never touches.
The step after. Coding happens in NVivo, ATLAS.ti, MAXQDA, Delve, or Taguette. A transcript that leaves as REFI-QDA or timestamped text saves the re-import work every round.
Where the audio goes. Ethics applications ask who processes the recording, in which country, whether it trains models, and whether a DPA is signed. The answer decides between local and cloud before price does.
The nine tools
Phonotheca
20 audio hours a month, searchable archive, REFI-QDA export
$12 a month
noScribe
Local Whisper with speaker labels, Windows and Mac
Free
MacWhisper Pro
Local Whisper as a Mac app, one-time licence
€65 once
Otter.ai Pro
1,200 minutes a month, 90 minutes per conversation, 10 imports
$16.99 a month
Happy Scribe Basic
120 minutes a month, 80+ languages
€17 a month
Rev (human)
Human-typed transcripts, per audio minute
$1.99 a minute
NVivo Transcription
Per-hour add-on to an NVivo licence
Add-on
Condens Lite
Unlimited transcription, one contributor
€15 a month
Dovetail
Team repository, Free and Enterprise plans
Free tier
Prices shown as of August 2026.
1. Phonotheca
A workspace for interview programmes rather than single studies. Upload audio, get speaker-labeled transcripts you can correct in place, and every interview joins one archive that search covers end to end. A project exports as a REFI-QDA .qdpx that opens in NVivo, ATLAS.ti, or MAXQDA, and processing stays in the EU on every plan with a data governance statement written for ethics applications.
Solo is $12 a month for 20 audio hours. The free plan transcribes 3 hours once and keeps the archive, search, and export working permanently. Doctoral work has its own version of this arithmetic, worked through in transcription on a dissertation budget. I build it, and the boundary is plain in the pricing: for a one-off study with no follow-up rounds, the two local tools below cover the job without a subscription.
2. noScribe
Free, open source, and local. It wraps Whisper and speaker diarization in a desktop app for Windows and Mac and exports formats the QDA tools read. Nothing leaves the machine, which settles most ethics conversations before they start. Budget roughly three hours of processing per hour of interview on an ordinary laptop, and expect to fix speaker labels by hand when the same participants return across rounds. aTrain covers the same ground on Windows.
3. MacWhisper Pro
The same local-Whisper approach as a polished Mac app, €65 once. It is faster than noScribe on Apple silicon and handles batches well, so it suits high volume on a fixed budget. Speaker labels take extra setup, and the output is a transcript file rather than a searchable corpus, so retrieval across a study stays your problem.
4. Otter.ai Pro
Built around live meetings, with recorded-file transcription attached. Pro is $16.99 a month for 1,200 minutes, and the limits that matter for fieldwork sit elsewhere: 90 minutes per conversation and 10 file imports a month, so a study of many short recordings hits the import cap with minutes to spare. The full comparison covers where each fits.
5. Happy Scribe
Subscription transcription with the widest language coverage on this list, 80+ languages and dialect variants. Basic is €17 a month for 120 minutes, extra minutes are €0.20 as top-ups, and human proofreading is available from €1.75 a minute. The per-hour cost sits at the high end for English interviews, and the language range is the reason to pay it. The full comparison covers retention and the DPA.
6. Rev
Human transcription at $1.99 a minute with a verbatim option, plus AI transcription at $0.25 a minute. When a committee or a publication requires every word right the first time, a typed-by-a-person transcript is what that costs. Files are bought one at a time and live in your downloads folder afterwards. The full comparison covers the archive side.
7. NVivo Transcription
The add-on path if your department already runs NVivo through a site licence. Transcription is sold separately per hour on top of the licence, and the transcript lands directly in the software the coding happens in, which is its argument. The licence stays tied to a machine and a version. The full comparison covers the trade.
8. Condens Lite
A research repository for teams, and the strongest deal on this list for a single contributor who wants unlimited transcription: Lite is €15 a month with no hour cap. The next tier is Business at €500 a month, which is where a growing team lands. The full comparison maps the two shapes.
9. Dovetail
A team repository with stakeholder-facing reporting, now sold as a free tier and an Enterprise contract with nothing between. The free tier is a way to evaluate the repository shape; a funded team talks to sales. The full comparison covers it.
Common questions
- What transcribes qualitative interviews for free?
- noScribe and aTrain, both of which run Whisper with speaker diarization on your own machine and export formats NVivo, ATLAS.ti, and Taguette read. Budget roughly three hours of processing per hour of interview on a normal laptop. Phonotheca's free plan transcribes three hours once and keeps the archive and search working after that.
- Should the limit be hours or files?
- Hours. Some tools cap uploads per month rather than duration, so a study of thirty short interviews hits a wall that a study of five long ones never touches. Check which shape the cap has before fieldwork starts.
- What do ethics boards ask about transcription?
- Who processes the audio, in which country, whether the vendor trains models on it, how long copies are retained, and whether a data processing agreement is signed. Local tools end the question because the audio never leaves the machine. For cloud tools, the data governance statement at /data-governance is written to be quoted in an ethics application.
- How does a transcript get into NVivo or ATLAS.ti?
- Cleanest is a REFI-QDA .qdpx archive, the interchange format both import along with MAXQDA and Dedoose. Failing that, plain text or SRT with timestamps imports everywhere, at the cost of re-linking quotes to audio by hand.
- How many audio hours does a study need?
- A typical interview study runs 8 to 15 interviews at 45 to 60 minutes, so 6 to 15 audio hours. Plans metered near 20 hours a month cover a study a month with room for re-recordings.
Related pages
Run the next study through it
Three audio hours are free, no card, and everything exports if you leave.