Transcription for qualitative research

Transcribe research interviews and focus groups without uploading them. Runs offline in your browser, labels the speakers, and exports for NVivo, MAXQDA and ATLAS.ti.

AudioRinse · transcriberOn-device · 0 B uploaded

Drop an interview here

MP3 · WAV · M4A · MP4 · MOV · WEBM — up to 3 hours. Nothing is uploaded.

The model downloads once and stays in this browser, so later transcripts start at once and work with Wi-Fi off. Your recording is never uploaded. Free: Fast transcripts without speaker labels, as text, SRT or VTT, no limit. Pro: Accurate, speaker labels, Word, NVivo and MAXQDA files.

How it works

  1. Pick Accurate, the number of speakers and the language. The model downloads once, then stays in your browser.
  2. Drop the interview or focus group onto the box above, up to 3 hours. Nothing is uploaded.
  3. Check it against the recording: click a line to hear it, name the speakers, correct words in place.
  4. Download it for NVivo, MAXQDA, ATLAS.ti or Word, with speakers and timestamps.

Ethics approval and data protection

Most ethics boards and data protection officers ask where recordings go and who can access them. With this tool the answer is short: nowhere and nobody. The audio is decoded and transcribed inside the browser on the researcher's own computer, and the transcript is saved only there. No account is needed to transcribe.

The data-flow statement is written for protocols and DPIAs: it lists each network request the page makes (the page itself, and the model files from Hugging Face the first time), and describes the Wi-Fi-off test. We can't approve a tool on your board's behalf, but you can quote or attach the statement.

Clean verbatim, and when you need more

Whisper writes clean verbatim: it leaves out most "um", "uh" and false starts, and keeps the words as spoken. That suits thematic analysis, grounded theory and most interview studies.

Conversation and discourse analysis need pauses, overlaps and fillers. Use the transcript as a first draft there: the timestamps and click-to-listen lines make it quick to add what your notation needs.

Into NVivo, MAXQDA and ATLAS.ti

  • NVivo: download NVivo, then import it as a transcript of the media file. Map Timespan, Content and Speaker; the file has a header row.
  • MAXQDA: download MAXQDA (or Word) and import it with the recording. The [hh:mm:ss] timestamps link each paragraph to its place in the audio.
  • ATLAS.ti: download VTT and add it as a transcript of the media document.
  • Anything else: Word keeps speakers and timestamps in each paragraph; Text is plain.

Accuracy on real interviews

On three recorded hour-long interviews with professional reference transcripts, Accurate got 8.3% of words wrong, on a par with desktop Whisper tools on the same files, and fewer than 2 words in 100 went to the wrong speaker. Accents and crosstalk cost the most. Budget a listening pass before you code, as you would with any automatic transcript.

Questions

Is the audio uploaded to a server?

No. Transcription and speaker detection run in your browser on your own computer, and the transcript is saved only in that browser. The only download is the model, once. The data-flow statement lists every request.

Can I use it under my ethics approval or GDPR?

Your board or data protection officer decides, but the facts they need are simple: the recording is processed on the researcher's device, nothing is transmitted, and it works with the network off. The data-flow statement is written to be quoted in a protocol or DPIA.

Does it work for focus groups?

Yes. Set the number of speakers, or choose "Not sure" to let it count them. With more than four people, or a lot of crosstalk, expect more lines to need a new speaker label.

Which formats can I export?

Word (DOCX), NVivo (tab-delimited with Timespan, Speaker and Content), MAXQDA ([hh:mm:ss] timestamps), VTT for ATLAS.ti and players, SRT subtitles, and plain text.

Does it work offline, for example during fieldwork?

Yes. After the first transcript the page and the model stay in your browser; with no connection it works the same. Your Pro plan is remembered for 30 days offline.

What does it cost?

Fast transcripts without speaker labels are free as text, SRT or VTT, with no limit. Pro is $7.99 a month or $39.99 a year and adds downloads of Accurate and speaker-labelled transcripts in every format. It's an individual plan; there's no per-hour fee, because the work happens on your computer.

Other tools