To transcribe an interview privately, open QuietSteno, choose the recording and select the spoken language. The speech model runs on your own device, so the audio is not uploaded to any server. You get a timestamped transcript that you can edit, label with speaker names, and download as a text file.
How to transcribe an interview
- Add the recordingDrag in the file from your recorder or phone: MP3, WAV, M4A or a video file.
- Choose language and accuracyUse Most accurate for interviews. Wording matters, and the extra time is worth it.
- Verify against the audioClick the timestamp beside any line to hear it. Check every quote you intend to use.
- Label speakers and exportType speaker names at each turn, then download TXT with your edits.
Why the location of processing matters
When you upload an interview to a cloud transcription service, a copy of it exists on that company’s systems, under its retention rules and within reach of its staff, subcontractors and legal process. For many interviews that is acceptable. For a whistleblower, a patient, a minor or a research participant who was promised that only the study team would hear the recording, it may not be.
With QuietSteno the recording is opened by your browser and processed by a model running on your own hardware. There is no account and no server-side copy, because the audio is never transmitted. The privacy page lists exactly what the site does load from the network.
Working with the transcript
- Timestamps as citations. Each line starts at a known time. Note the timestamp next to a quote in your draft and you can find it again in seconds.
- Verbatim or clean? The model writes what it hears but tends to drop some fillers such as “um”. If you need strict verbatim for discourse analysis, add those by ear.
- Quotes need checking. Automatic transcription makes mistakes with names, numbers and negatives. Treat the transcript as a draft and confirm anything you quote against the audio.
- Qualitative analysis. The TXT export opens in tools such as NVivo, ATLAS.ti and MAXQDA, or in any word processor.
Questions and answers
Is this suitable for confidential or sensitive interviews?
The audio is processed on your device and not uploaded, which removes the third-party copy that cloud services create. You are still responsible for the security of your own computer and for any rules set by your institution or ethics board.
Does QuietSteno label who is speaking?
No. The transcript is one continuous text with a timestamp on every line. Each line is editable, so you can type speaker names in front of the lines before you copy or download.
How do I check that nothing is uploaded?
Open your browser’s developer tools, switch to the Network tab and run a transcription. You will see the model files being downloaded on first use and no request carrying your audio. You can also disconnect from the internet once the model has loaded; transcription continues.
Does it handle two people talking over each other?
Not well. Overlapping speech is hard for every speech model. It will usually capture the louder voice. Mark those passages and fix them by ear.
Is there a limit on interview length?
QuietSteno sets no limit on file count, length or minutes, and there is no account. The practical limit is your device: the browser has to hold the decoded audio in memory, so a recording of several hours can fail on a phone or an older laptop. If that happens, split the file into parts.
Related
- Meeting transcriptionZoom, Teams and Meet recordings.
- Audio to textMP3, WAV, M4A, OGG or FLAC to text.
- Is online transcription private?Where your recording goes, and how to check.
- Choosing an accuracy settingWhat Fast, Balanced and Most accurate mean.
Last updated 13 November 2026.