Terms for the voice transcript
Version of 19 November 2026
Translation: this English version is provided to help you understand the terms for the voice transcript. The language of the contract is German, and only the German version is legally binding: https://parafeld.de/sprachtranskript-bedingungen/
These terms apply to the voice transcript in the “Documentation” widget of the Parafeld™ System Board. Part A supplements the terms of use (hereinafter “main contract”), and Part B supplements the data processing agreement. If you do not use the voice transcript, you do not need to accept these terms. The main contract and the data processing agreement continue to apply unchanged; these terms are added to them for the voice transcript.
The provider is METAFELD · Ulrich Reuther, Zossener Straße 41, 10961 Berlin, Germany, email service@parafeld.de (hereinafter “we”).
Part A – Supplement to the main contract
A 1 What the voice transcript is
(1) With the voice transcript you record audio in the board through the microphone of your device: as a dictation, in which you speak on your own, or as a recording of a session. The audio is converted into text on our own servers at Hetzner Online GmbH in Nuremberg with the freely available speech recognition model Whisper and then deleted. The text appears as a transcript in the documentation of the constellation; you can change it, copy it, add it to the session notes or the report, and delete it.
(2) The following are not part of the voice transcript: recognising who is speaking; a summary or analysis of the conversation; keeping the audio recording; capturing the audio of other programs, such as a video conference. What is recorded is what the microphone of your device picks up.
(3) The transcript is your private documentation, like the session notes. It does not go into the live session, into the scene link or 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, not end-to-end encrypted.
(4) If you record during a live session, everyone taking part sees a notice that the facilitator is recording the audio. In a live session, only the facilitator can record.
A 2 Requirements, costs and limits
(1) To use the voice transcript you need an account and must be signed in. Before your first recording, you accept these terms in the board.
(2) During the introductory phase the voice transcript is free of charge. What applies later: like the folders, the voice transcript will then be part of the Practice plan. When the plan starts, accounts from the introductory phase get a trial period of four months. Nothing becomes chargeable on its own – nobody pays without ordering it themselves. Transcripts you have already received remain in your constellations. We will announce the start by email and on the Pricing page.
(3) Each account can record at most 20 hours in 30 days, calculated over the most recent 30 days in each case, at most 3 hours per session recording, at most 30 minutes per dictation, and one recording per account at a time. What counts is the time during which a recording is running; a pause counts for at most five minutes. The parts of a recording are converted in the order in which they arrive. If many recordings are running at the same time, a new recording is still accepted; the conversion then takes longer, in individual cases several hours, and the board tells you the expected waiting time. There is no commitment to a particular processing time. Only when demand is exceptionally high may it be temporarily impossible to start a new recording; the board says so before the start.
A 3 Your obligations
(1) You record a session only if all persons who speak or may be heard in it know about the recording beforehand and have consented. This also applies to persons who join later and to persons whose voice reaches the microphone through a loudspeaker. Before each recording of a session, you confirm in the board that everyone present knows about it and agrees.
(2) Please note: anyone who records the privately spoken words of another person without authorisation may be committing a criminal offence under Section 201 of the German Criminal Code (§ 201 StGB). In addition, the content of conversations in therapy and counselling is specially protected – by the duty of confidentiality of many professions and as health data or other special categories of personal data under Art. 9 GDPR.
(3) You are responsible for the consent of the persons recorded, for informing them and for the legal basis of the recording. Whether and how you use recordings in your work is for you to assess under your professional regulations; responsibility for professional regulations and confidentiality remains with you.
(4) The transcript is a machine conversion. It can contain errors – words, names or numbers recognised incorrectly, missing passages – and does not distinguish who is speaking. Check it before you use it. It does not replace any documentation that your profession requires.
(5) We recommend recording only what you need for your work and deleting transcripts as soon as you no longer need them.
A 4 Availability and changes
(1) We do not commit to any particular availability or any particular processing time for the voice transcript. If our server cannot be reached at the moment, the board keeps the audio in the open window and sends it again later. If the window is closed before a part has reached us, that part is lost; the board warns you beforehand.
(2) We may change or discontinue the voice transcript. Transcripts you have already received remain in your constellations.
(3) If we change these terms, the new version applies only once you accept it. The board asks you to do so before your next recording; without your acceptance, no new recording starts.
(4) In all other respects the main contract applies, including with regard to liability.
Part B – Supplement to the data processing agreement
This part applies if, as a business (Unternehmer, § 14 BGB, German Civil Code), you record data of other people with the voice transcript, for example of clients. It supplements the data processing agreement (hereinafter “DPA”); terms are used as defined there, in particular “controller” and “processor”. In all other respects the DPA applies. In the event of contradictions, this part takes precedence for the voice transcript.
B 1 Subject matter and purpose
(1) The processor converts audio recordings that the controller starts in the board into text, holds the text ready until the controller’s browser has collected it, and then deletes audio and text on its servers.
(2) Nature of the processing: receiving the audio in parts, passing it on to the conversion server, converting it into text, storing it briefly, transmitting it to the controller’s browser, deleting it. In addition to the DPA, instructions are the controller’s actions in the board: starting, pausing and ending a recording, cancelling and deleting.
(3) Transcripts that the controller saves with a constellation in a folder belong to the folders and constellations under section 12 of the DPA; the DPA applies to them directly.
B 2 Types of data and data subjects
(1) Data: audio recordings with the voices and everything that is said; the text generated from them, with time stamps; details of the recording (type, language, start, duration, number of parts). Depending on the conversation, audio and text contain health data or other special categories of personal data (Art. 9 GDPR), for example on mental health, and confidential information about family and relationships. The controller decides what is recorded.
(2) Data subjects: clients; other persons present, such as relatives, representatives or participants in a group; persons who are talked about.
B 3 Locations and sub-processors
(1) Audio and text are processed exclusively on two of the processor’s own servers at Hetzner Online GmbH in the Nuremberg data centre (Germany): on the application server (api.parafeld.de) and on a conversion server. No processing takes place outside Germany; audio and text are not transmitted to third parties.
(2) Hetzner Online GmbH is already approved under section 14 of the DPA; its service additionally covers the conversion server. No further sub-processors are added.
(3) The speech recognition model Whisper is freely available software and runs on the processor’s own server. The maker of the model receives no data and is not a sub-processor.
B 4 Periods
| What | Where | How long |
|---|---|---|
| Audio during the recording | working memory of the browser | until the part has reached the application server and has been accepted |
| Audio on the application server | in working memory only, not as a file | for the time it takes to pass it on to the conversion server |
| Audio on the conversion server | as a file | until the part has been converted (usually a few minutes, longer at busy times, about one day at most), then deleted; not included in the nightly backup |
| Text on the conversion server | database of the conversion service | until the application server has collected it – usually within seconds, otherwise within about a day |
| Text on the application server | separate file alongside the account database, not included in the nightly database backup | until the controller’s browser confirms receipt, no longer than 7 days after the recording began |
| Transcript in the constellation | the controller’s browser; after saving in a folder, on the application server like the rest of the constellation | until the controller deletes it (section 9 of the DPA) |
If the controller cancels a recording or deletes their account, the processor deletes the audio and text of these recordings on both servers without delay; a part that is being converted at that moment is deleted straight after its conversion. If a server cannot be reached at that time, the periods in the table apply. Deleted data disappears from the backups and the images of the servers when these expire, after 30 days at the latest (section 9 (2) of the DPA).
B 5 Additional technical and organisational measures
| Area | Measure |
|---|---|
| Audio | on the application server in working memory only; on the conversion server as a file with a neutral name, only until the conversion is complete, excluded from the nightly backup |
| Text | on the application server in a separate file outside the database backup, only until it is collected; the job at the conversion service is deleted straight after collection |
| Deletion | deleted entries in the databases of the voice transcript are overwritten; clean-up every hour according to the periods in B 4 |
| Separation | each recording belongs to one account and can be reached only when signed in to that account |
| Logs | no content in the logs – neither audio nor text, only identifiers, numbers and durations |
| Encryption | transfer between browser and servers and between the two servers only encrypted (TLS); on the servers, audio and text are unencrypted while they are being processed |
| Live session | the server learns only that the facilitator is currently recording (for the notice to everyone taking part); this information is held only in working memory |
In all other respects the measures under section 13 of the DPA apply.
B 6 Support with requests
Transcripts are held by the controller: the controller can view them in the board, correct them, export them with the constellation and delete them. The processor does not keep any audio recording. Requests from data subjects that the processor receives are forwarded to the controller in accordance with section 6 of the DPA.
Acceptance and version
(1) You accept these terms in the board before you dictate or record a session for the first time, with the box “I agree to the terms for the voice transcript.” We record your acceptance with your account, the version and the time. Part A and, insofar as you act as a business, Part B are thereby agreed electronically (Art. 28 (9) GDPR). On request we will confirm this in text form.
(2) The language of the contract is German. Versions of these terms in other languages are provided to help you understand them; the German version is authoritative.
How we handle your own data when you use the voice transcript is described in the privacy policy.