Use case

Research Interview Transcription

Get first-draft transcripts of one-to-one research interviews for coding in NVivo, ATLAS.ti, MAXQDA or a spreadsheet, with timestamps for checking quotes. The transcript doesn't name speakers, so you mark Interviewer and Participant yourself; with two alternating voices that's quick. Audio is deleted after delivery and never used for model training.

01 Upload

Your audio

Audio and text outputText only — timestamped transcript with no audio in results.
Server assistOff by default. When on, this job uploads your media file so our servers can decode it (helpful on phones). Uploads this file (max 500 MB). Same credit cost as a normal run. Max length still 2 hours.

By default the audio file stays in this browser (only processed audio chunks go to the cloud). Default is auto-detect across 100 languages; pick a spoken language to pin it, or one of 21 languages to translate.

02 Result

Your transcript

Your timestamped transcript will appear here.

Waiting for upload

105 languages available

Why it helps

Ethics boards ask where participant audio goes. mydubly's answer is short: audio is processed on our servers, deleted after delivery (within 30 minutes at most), and never used for training. The account only keeps the file name, length and credits.

A machine transcript is a first draft. Researchers still review it against the audio, but that's far faster than typing from scratch.

Step by step

  1. Rename files with participant codes (P07-session1.wav), not names: the file name is the one thing stored with your account.
  2. Transcribe in text-only mode; add a target language for interviews in another language.
  3. Import the timestamped transcript into your analysis tool and review it against the recording while coding.

Tips

  • Mention the service in your data-management plan and consent form as an automated transcription processor.
  • Record each voice clearly: a lapel or table microphone near the participant helps more than any setting.

Example

58 minutes

A PhD student transcribes twelve 58-minute interviews for 696 credits in total, under 70¢.

Cost: 290 credits (29.0¢).

Limitations to know

  • No speaker labels: fine for two alternating voices, slow for group discussions.
  • Verify transcripts before quoting participants.

Built for one-to-one interviews

Two voices taking turns, interviewer and participant, are the case this fits best. The transcript comes back as continuous text with timestamps and no speaker names, so you add "I:" and "P:" while you review; with two alternating voices that takes a few minutes per interview. Focus groups and group interviews are harder: several voices overlap, nobody is named, and assigning lines to participants means listening closely throughout. If your study is mostly focus groups, budget extra time or use a tool that labels speakers. Focus group transcription covers what that work involves.

Wording for your ethics application

Ethics committees want to know where participant audio goes and who can access it. A description you can adapt:

Data-flow statement

Interview recordings will be transcribed with mydubly, an automated transcription service. Audio is sent over HTTPS, processed to produce a transcript, and deleted together with the transcript within 30 minutes of delivery. Recordings are not used to train models. The service keeps only the file name, duration, languages and credits used, so files are named with participant codes. For video interviews, the picture is not uploaded; only the audio track is sent.

Check the statement against your institution's rules on external processors, and mention the service in the consent form and data-management plan. AI transcription and research ethics lists the questions committees most often ask.

Interviews in another language

Fieldwork interviews are often in a language your supervisor or co-authors don't read. Choose a target language and the job returns the original-language transcript and a translated transcript side by side, from the same upload and for the same price as a plain transcript. The usual advice holds: code and quote from the original, and use the translation so the rest of the team can follow and check your interpretation. Translating qualitative interviews explains how to handle meaning that doesn't carry across.

Do you have to transcribe every interview?

It depends on your method and your supervisor. Thematic and discourse analysis usually work from full transcripts; some applied studies code from detailed notes plus selected transcribed passages. Decide on verbatim or clean verbatim before you start: the draft drops most fillers like "um", so add them back where your method needs pauses and hesitations. Verbatim versus clean verbatim and transcription conventions in research set out the options.

Into NVivo, ATLAS.ti, MAXQDA or a spreadsheet

  1. Download the plain transcript for coding and the timestamped transcript for checking.
  2. Add speaker markers and remove anything identifying, such as names and places, replacing them with codes.
  3. Import the text file into your analysis software as a document. There is no direct integration; the files are plain text, which these programs accept.
  4. Keep the timestamped version beside it, so any coded passage can be traced to its moment in the recording.

Checking the draft before you quote

Read each transcript once against the audio, concentrating on quotes you might use, numbers, names and technical terms. Mark words you can't resolve as inaudible rather than guessing. Some researchers send transcripts back to participants to confirm; member checking describes when that helps. Proofreading an AI transcript gives a quick routine.

When an offline tool is the better fit

Some committees require recordings to stay on the researcher's own computer. If yours does, use an offline transcription tool that runs a speech model locally; several open-source options built on Whisper exist for exactly this. mydubly processes audio on its servers, so it fits only where approved external processing is allowed.

Costing a whole study

Twelve 58-minute interviews
696 credits (69.6¢) as plain transcripts
The same twelve with English translations
Another 696 credits (69.6¢) if the interviews are in another language and you run them with English as the target
Free credits
100 credits with Google sign-in, enough for one or two interviews

Run one interview on the free credits as a pilot, check the transcript against the recording, then top up for the rest of the study from $5.

Using the tool on this page

  • The spoken language is detected automatically across about 100 languages; you can pin it, or pick one of 21 languages to translate into. Files can be up to 2 hours long.
  • Audio is sent over HTTPS in small chunks and deleted with its transcripts within 120 minutes of delivery (how files are protected).
  • You can download a timestamped transcript, subtitles (SRT and VTT) and plain text. Every output, step and limit is explained on the audio to text page.

Frequently asked questions

Is it suitable for sensitive data?

Audio is encrypted in transit, processed and deleted after delivery, and never used for training. Check your institution's rules for third-party processors before using any external service.

Can I export to NVivo?

The transcripts are plain text files, which NVivo, ATLAS.ti and MAXQDA import directly.

Can it tell the interviewer and participant apart?

No. Transcripts don't carry speaker names. With two alternating voices, adding I: and P: markers during review is quick.

Can I transcribe interviews in another language and get English too?

Yes. Choose English as the target language and you get the original-language transcript and an English translation from the same upload.

Is it suitable for a dissertation?

As a first draft, yes, for one-to-one interviews where approved external processing is allowed. Review every transcript against the audio before quoting, and follow your supervisor's conventions.