Transcription for a whole study

The work between fieldwork and coding

1. Fieldwork

   Upload the whole round

   One project holds one study. Add interviews as you run them, from a handheld recorder or a phone.

2. Transcript

   Speakers labeled, turns timestamped

   Interviewer and participant are separated automatically. Where the split is wrong, cut the segment and reassign the speaker yourself.

3. Familiarization

   Mark passages on the first read

   Tag while the interview is fresh. The highlight stays on the passage, with the timestamp that points back to the audio.

4. Write-up

   Quote with the citation attached

   Copy a passage and its participant, interview, and timestamp come with it, ready to drop into a chapter.

Common questions from research teams

I remember a participant saying something but not who or when. Now what?

Search every transcript in the study at once. Each result carries the participant, the interview, and the timestamp, and one click plays that point in the audio. It works whether or not you tagged the passage at the time.

My supervisor says transcribing by hand is part of the analysis.

Then correct instead of type. The transcript comes back with speakers and timestamps and you work through it against the audio, splitting segments and reassigning turns where they are wrong. That pass is the familiarization, and it ends with a text you can search for the rest of the project, which a recording never is.

Why not run Whisper or MacWhisper on my own machine?

For a dozen interviews you read once, a local tool does the job. What it leaves you with is a folder of files on one laptop. Three things usually decide it. Whether you interview the same people again, because speaker names have to stay consistent across a project. Whether the study runs past twenty or thirty hours, because that is where the setup and the waiting stop being free. And whether a supervisor or a colleague needs to open the material.

Does it code data the way NVivo does?

It does light tagging: define your themes per project, mark passages, and filter or export by tag. For framework matrices, memo networks, and inter-coder agreement, export the project as REFI-QDA and code it in the suite you already use.

Which coding software can open the export?

The REFI-QDA .qdpx archive imports into NVivo, ATLAS.ti, MAXQDA, Dedoose, and Quirkos. Transcripts arrive as sources and your tags as codes, with each coded passage still attached to its stretch of text.

What can I put in an ethics application?

The data governance statement is written to be quoted or attached. It names where processing happens, lists every sub-processor with its country, sets out retention and deletion, and ends with two paragraphs you can paste straight into a consent form or an ethics application.

My participants have strong accents or speak English as a second language.

Set the language on the recording rather than leaving it to detection, which is where heavily accented speech is most often misread. Speaker separation holds independently of how cleanly the words come out, so turns stay attributed even where wording needs fixing, and any passage can be corrected against the audio without retyping the interview.

How accurate are the speaker labels on a two-person interview?

Interviewer and participant separate cleanly on most recordings. Short interjections sometimes get absorbed into the other speaker's turn, so any segment can be split at a word and reassigned, and tags follow both halves.

I already had some interviews transcribed. Can they go in?

Import them as TXT, SRT, or VTT and they join the project as searchable, taggable turns with their speakers and timings intact. Imports cost no transcription hours, so earlier rounds of a study sit next to the ones we transcribe.

What happens when my hours run out?

New transcription pauses. Everything already in the archive stays readable, searchable, and exportable, including on the free plan.

Compared with your current tool