Interview transcripts ready to import into NVivo
Upload a research interview and get a transcript with a speaker label on every paragraph, as Word and TXT documents, plus a CSV with one row per speaker turn: File, Time, Speaker, Text and Translation. NVivo, ATLAS.ti and MAXQDA can import either form. Up to 15 minutes is free in full; a longer interview is $5, up to 6 hours.
- CSV columns: File, Time, Speaker, Text, Translation; one row per paragraph
- Every paragraph starts with a speaker label you can rename
- Word, PDF, TXT and SRT or VTT in the same download
- A neutral summary, topics in order and notable quotes; no coding by the AI
- Optional translation under each paragraph and in its own CSV column
How to prepare an interview transcript for NVivo
-
1
Upload the interview
Add the recording, or paste a public Google Drive or Dropbox link. If one session was saved as several files, add them together in name order.
-
2
Set the speaker labels
Use Find and replace on screen to turn "Speaker 1" and "Speaker 2" into Interviewer and a participant code.
-
3
Download and import
Pay $5, download the Word file to import as a document or the speaker-turn CSV to import as a dataset.
Two ways into your coding software
Most researchers import interviews as documents: one Word or TXT file per participant, coded line by line. The Word download works for that, with a speaker label at the start of each paragraph and a timestamp you can use to go back to the audio.
The CSV suits a different workflow. Each row is one speaker turn, so it can be imported as a dataset, where columns become fields you can sort and filter: all turns by the participant, all turns after the first hour, every turn next to its translation. NVivo, ATLAS.ti and MAXQDA each have an import for spreadsheet or survey-style data; the steps differ between versions, so follow your program’s help on importing a dataset from Excel or CSV. Excel and Google Sheets open the CSV directly if you want to look first.
Rename speakers before you export
Speakers are labeled "Speaker 1", "Speaker 2". Use Find and replace on screen to change them to "Interviewer" and "P07", or whatever your protocol uses, before downloading, so every file carries the same labels. Real names appear only if they are said in the recording. Speaker numbers are assigned per ten-minute part, so on a 90-minute interview check that the interviewer kept the same number throughout.
What the AI does and does not do
- Writes every word in the language spoken, with [inaudible] where speech is unclear
- Gives a neutral summary and a one-sentence neutral summary per topic or question
- Picks notable quotes word for word, with speaker and time
- Does not code, tag themes or interpret what participants meant
- Does not time pauses or mark overlapping speech in detail, as conversation analysis requires
Ethics and data
Check that your ethics approval and consent forms allow transcription by an outside service. The recording is visible only to you, sent to an AI model provider only to be processed, and not used to train models. Jobs started without an account are deleted after 7 days, and you can delete a job at any time. The transcript is a draft made by an AI speech model, so check quotes against the audio before you publish them.
How the tool works, what it costs and how your files are handled is on the Interview Transcription page; prices for every tool are on the pricing page.
Last updated
Questions and answers
Can ChatGPT transcribe an interview for NVivo?
For a short excerpt it can produce usable text. A full interview of an hour or more usually exceeds the audio length limit, and a long transcript gets shortened, which leaves gaps you cannot code.
I have 20 interviews. Is there a bulk option?
Yes. Bulk audio transcription takes up to 100 files for one payment of $5, up to 6 hours in total, and gives a TXT per interview. It does not produce the speaker-turn CSV.
What does one interview cost?
$5, paid once, for an interview of up to 6 hours. Up to 15 minutes is free in full. Unlimited, $19 a week, suits a week of fieldwork.
Does the CSV work with ATLAS.ti and MAXQDA?
It is a plain CSV, so any program that imports spreadsheet data can read it. Check your version’s help for the exact import steps.