Qualitative interview transcription

Export DOCX and import it as a document source. All three programs read it directly, and timestamps in the text survive the trip so you can jump back to the audio moment a code refers to. Plain TXT works too if your project prefers minimal formatting.

Edit identifying names directly in the transcript before exporting, and the corrections propagate to every export. Renaming diarized speakers to "P1", "P2" takes seconds, on any plan.

Accents transcribe well with modern models. For jargon-heavy interviews, add your key terms (drug names, theory terms, place names) to the custom vocabulary box before uploading, so they come out right instead of being guessed at phonetically.

It depends on the method. AI output lands close to clean read: false starts and fillers are tidied into readable prose, which suits thematic analysis and most coding work. If your approach treats hesitations and repairs as data (conversation analysis, for instance), do a listen-and-restore pass on the passages you code rather than trusting any automated judgment about what counts as disfluency.

Most committees accept it when the researcher verifies transcripts against the audio and the data handling is documented. The specifics an IRB tends to ask about are covered here: encryption in storage, the 30-day auto-delete of recordings, and no training on your data. Put those into your data-management plan and the method section becomes a one-line description plus a verification statement.

If interviews average an hour, that's ~40 hours of audio. Pro ($19/month, 20 hours per 30 days) covers it across two cycles, or Business ($79/month, 100 hours) in one. Either way it is pennies against a transcription bureau's $1.50+/minute.