Saypane
English

Transcribe research interviews privately: a local workflow

Consent forms, ethics approvals, GDPR assessments and hospital policies often say where interview recordings may go. Online transcription services usually do not fit. Local transcription keeps the audio on the computer where you already store it. This guide explains what that means in practice and how to set it up.

Not legal advice. Whether a tool meets your obligations is decided by your ethics board, data protection officer or compliance team. This page gives you the facts to bring to them.

What "local" means with Saypane

  • Speech recognition runs on your PC. The audio and the transcript are not uploaded to us or to anyone else.
  • The app has no account, no analytics and no tracking.
  • Saypane uses the network only to download speech models and for Microsoft Store purchases and updates. Model downloads send no audio and no text.
  • Once the models are downloaded, you can disconnect the PC from the network and transcribe as usual. That is the simplest way to show that nothing leaves the machine.

Because no outside service processes the recordings, there is no transcription provider to name in a data processing agreement or a consent form. The usual rules for storing the files still apply.

A workflow that keeps everything on one PC

  1. Download the models first, on a normal connection: open Saypane and pick your interview language so the model is fetched.
  2. Copy the recordings from the recorder or phone to an encrypted drive on your PC. On Windows, BitLocker encrypts the whole drive.
  3. Optional: go offline. Turn off Wi-Fi or unplug the cable while you transcribe.
  4. Transcribe each interview, or queue them all with Pro. Turn on speaker labels for two or more voices.
  5. Check the transcript against the audio. Fix names, technical terms and speaker labels.
  6. Pseudonymize in the exported file: replace names and identifying details as your protocol requires.
  7. Export to DOCX for Word or your analysis software, and store it next to the audio.

Verbatim, not summarized

Saypane writes down what was said, in order, with timestamps. It does not summarize, paraphrase or "clean up" the content with an AI model. Two things to know:

  • Speech models tend to leave out filler words ("um", "uh"), repeated words and false starts. For thematic analysis that is usually fine. For conversation analysis or other methods that need every hesitation, add them while you check the transcript.
  • Unclear passages may be transcribed as something plausible but wrong. Check every quote you use against the audio.

Speaker labels

With Saypane Pro, each passage is labeled Speaker 1, Speaker 2 and so on. That separates interviewer and participant, and the voices in a focus group. Labels are worked out from the voices and are not always right: two people talking at once, similar voices or a single microphone far from everyone lead to mistakes. Review them before analysis.

Long interviews

Saypane splits long recordings at pauses and transcribes each part separately, checking for the repeated-sentence loops that Whisper is known to produce on long files. A 90-minute interview is handled the same way as a 5-minute one. With Parakeet on a 4-core CPU, an hour of English, German, French, Spanish or another of its 25 languages took under 5 minutes in our test.

What to tell your ethics board or DPO

  • Software: Saypane for Windows, from the Microsoft Store.
  • Processing: on the researcher's PC. No cloud service, no account, no analytics.
  • Network use: model downloads (from models.saypane.com or Hugging Face) and the Microsoft Store only. Transcription works with the network disconnected.
  • Storage: where the researcher saves the files, for example an encrypted drive.
  • Details: our privacy policy.

Questions

Is local transcription GDPR compliant?

Compliance depends on your whole process, not on one tool. Local transcription removes one common problem: no third party processes the recordings, so there is no transfer to a transcription provider to justify. Storage, access and deletion of the files are still your responsibility.

Can I try it on a real interview before buying?

Yes. The free version transcribes the first 15 minutes of every file, completely and without a watermark, so you can judge the accuracy on your own recordings. Speaker labels need Saypane Pro.

Last updated October 27, 2026. Still stuck? Contact support.

Other guides