Voice transcript: sessions and dictation as text
Record your own impressions after a session or record the whole conversation: the Parafeld™ System Board converts the audio into text and adds it as a transcript to the documentation of the constellation. The transcription takes place on our own servers at Hetzner Online GmbH in Nuremberg, Germany; the audio is deleted afterwards.

Two ways: dictate or record the session
The voice transcript is a section in the “Documentation” widget, below the session notes, with two buttons:
- Dictate – you speak on your own, before or after the session: observations, hypotheses, sentences that stayed with you, ideas for next time.
- Record session – the whole conversation is recorded. This is only allowed if everyone present knows about it and agrees. Before the start you confirm this with a tick; without the tick, no recording starts.
How a recording works
- Start A short dialog explains what happens to the audio. There you choose the language of the recording – German, English, French, Spanish or “Detect automatically”. The first time, you accept the terms for the voice transcript, and the browser asks whether the site may use the microphone.
- Record While the recording is running, an indicator – on a computer at the top right of the header, on a phone above the bar at the bottom – shows a red dot, the elapsed time and the microphone level, together with “Pause” and “Stop”. In a live session only the facilitator records, and everyone taking part sees the notice “The facilitator is recording the audio of this session”.
- Receive the text While the recording is still running, the audio is sent to the server in parts of a few minutes each and converted part by part. That is why the transcript usually appears in the list soon after “Stop”. If you close the board before then, it collects the text as soon as you open it again and are signed in – within seven days of the start of the recording.
Transcription on our own servers in Germany
The audio is sent encrypted to our server api.parafeld.de and from there to a second server of our own, on which the speech recognition model Whisper converts it into text. Both are located at Hetzner Online GmbH in Nuremberg. Whisper is a freely available speech recognition model that runs on our own server; no data goes to the maker of the model or to any other third party, and nothing is processed outside Germany. We do not use audio or text for purposes of our own, nor to train AI models.
- The audio is not kept: once a part has been converted – usually after a few minutes – it is deleted. The audio is not included in the nightly backup.
- The text remains on our servers only until your browser has collected it, no longer than seven days after the recording began.
- The transcript then stays in the constellation until you delete it.
All periods in detail are listed under Security and privacy.
What the transcript is – and what it is not
The transcript is a machine conversion of what was said. It does not distinguish who is speaking; for a session, each paragraph begins with a time stamp such as [12:34]. Names, technical terms and anything spoken indistinctly may be recognised incorrectly. So read a transcript through before using it; it can be corrected in the text field. The board does not write a summary, and you do not get the audio recording itself back.
Only what the microphone of your device picks up is recorded, not the audio of other programs. In a video session with headphones, this is usually only your own voice.
How it helps in therapy and counselling
- Capture impressions straight after the session, without typing.
- Re-read important sentences word for word – a resolving sentence, the words of a representative, a turning point.
- A basis for your own documentation, such as therapy or coaching notes: “Copy” puts the text on the clipboard, “Add to session notes” appends it to the notes, and with the “Include in report” tick box it appears in the report as a PDF.
The transcript does not replace any documentation that your profession requires. What part of it belongs in your records is for you to decide.
What stays private
Transcripts are your own documentation, like the session notes: they never go into the live session, never into a scene link and never into the picture to take home. If you save the constellation in a folder in your account, the transcript is saved with it – like the rest of the constellation, unencrypted on our server, that is, not end-to-end encrypted.
What you need
- An account: you need to be signed in (Create account). During the introductory phase the voice transcript is free of charge. Later it will be part of the Practice plan, like the folders; nothing becomes chargeable on its own. More under Pricing.
- An up-to-date browser with a microphone on a computer, tablet or phone; before the first start, the “I’m not a robot” check, as when saving in folders.
- Limits: 20 hours of recording per account in 30 days, at most 3 hours per session, at most 30 minutes per dictation, one recording at a time. If many recordings are in progress at once, the board still accepts yours: the parts then wait in a queue, the transcript arrives later, and the board tells you the expected waiting time.
The individual buttons are described on the page “Documentation” widget (in German).