DocumentationTools
Live transcription and meetings
Transcription and Meetings turn speech into text — live, via upload, or through the public API. Speaker labels, anonymization and meeting notes are optional. Afterwards you continue in chat or Klartext.
5 Min
Where the features live
- Apps → Transcribe / Meetings: start a recording or import an audio file.
- Apps → Klartext: desktop dictation and imported plain transcripts (text view without meeting tabs).
- API: POST /api/v1/transcriptions (async 202) — same engine as the app; legacy alias POST /api/transcribe.
Two modes: transcript and meeting
| Mode | Result | Limits |
|---|---|---|
| transcript | Text only. Speakers optional. No automatic notes, no meeting title. | Up to 100 MB |
| meeting | Text plus title and default summary; speakers on by default. | Up to 250 MB |
Formats: MP3, WAV, M4A, OGG, WEBM. In Meetings and Klartext, Import audio file starts the same ingest: Meetings create a meeting job; Klartext creates a transcript job (source=klartext).
Typical flow
- 01Start recording or import / upload audio.
- 02Wait for transcription (processing/pending → completed). Speakers and timestamps when requested.
- 03Review the text. Anonymize if needed before it goes into a chat or knowledge base.
- 04In chat, summarize, extract tasks or export as an artifact. Meeting jobs open the meeting view; plain transcripts open the Klartext detail page.
Anonymization
For sensitive client conversations: anonymize before the transcript hits an external frontier model. In the API, anonymize=true runs the same pipeline; GET then returns the redacted text without original segments. The dictionary applies here too.
Difference from Klartext
Meetings and live transcription are for conversations, files and minutes. Klartext is desktop voice input into any program (hotkeys, Protected mode) and the text view for imported plain transcripts — without meeting list, Sachkern, Analyse or speaker assignment.