Editing transcripts

Clicking an entry in the Transkripte overview opens the detail view. There you see the recording’s metadata, can play the audio, correct the recognised text, name the speakers, and export the result.

Recording metadata

The header of the detail view lists the technical details of the transcription.

Field	Content
Status	In Bearbeitung (in progress) while transcribing, then Abgeschlossen (completed)
Erstellt	When the recording was created
Abgeschlossen	When the transcript was finished
Verarbeitungsdauer	Duration of the transcription run
Audiolänge	Length of the recording
Erstellt von	The user who owns the transcript
Sprache	Language of the recording
Sprechererkennung	Aktiv or Aus, depending on the setting used when recording

While transcription is running, the note “Wird transkribiert… Je nach Länge der Aufnahme kann dies einige Minuten dauern.” (Transcribing… depending on the length of the recording this may take a few minutes) appears in place of the text. Bearbeiten (Edit) in the header changes the entry’s title and description.

Note

In the example above the Sprache field shows a dash, even though the recording was in German. Whether and when the language is populated — and whether it can be selected when creating a recording — is not visible in the underlying recording and should be verified before publication.

Playing the audio

An audio player sits under Audio, so the recording can be replayed at any time — useful for checking an unclear passage in the transcript against the original.

Correcting the text

The Transkript section offers two actions:

• Bearbeiten (Edit) — switches the text into an editable field. The action then changes to Fertig (Done), which commits the change.
• Synchronisieren (Synchronise) — realigns text and audio.

Typically you correct proper nouns and technical terms the recognition misheard here — for example an “App” that should read “API”. The correction also carries through to the export.

Naming speakers

If the recording was made with speaker diarization active, a Sprecher (Speakers) block appears above the text with one input field per detected voice (Sprecher 1, Sprecher 2, Sprecher 3, …).

Enter the actual names there. The assignment then carries through the whole transcript: each contribution is headed with the name you gave and a timestamp. What was an anonymous numbering becomes a readable record of the conversation — and the names also appear in the exported file.

Note

Speaker diarization distinguishes up to 35 voices per recording. Name the speakers once centrally in this block rather than overwriting individual sections in the text.

Exporting as Markdown

The .md herunterladen button in the top right saves the complete transcript as a Markdown file. With multiple speakers, the file carries the names you assigned as headings, so it can be processed further right away — as the basis for minutes, for example, or for filing in a knowledge collection.

What next?

Instead of processing the transcript by hand, you can have it analysed directly in the chat: Access from CompanyGPT.