Skip to main content

Meeting Transcription

Harmony turns meeting audio into a full, timestamped transcript with speaker labels. You can edit it, search it, add custom vocabulary for better accuracy, and export it in several formats.

What you get​

Every processed meeting has a transcript showing who said what and when. Each segment has a timestamp so you can jump to a moment, create clips, or share a specific part.

Use search (Ctrl/Cmd + F) inside the transcript to find terms and move between matches. Harmony identifies speakers from meeting attendees and contacts when possible. Unknown speakers appear as "Speaker X".

Correcting or reassigning speaker labels is not available yet and is under development.

Editing the transcript​

Open the meeting and go to the Transcript tab. Click any text to edit. Changes save automatically.

Correcting names, companies, and technical terms helps the most. Harmony uses your corrections to improve future transcripts. You can also add terms in Workspace Settings → General → Handbook so they are recognized from the start.

Speaker labels​

Harmony assigns speaker labels using meeting attendees and your contacts. When it cannot identify someone, they appear as "Unknown" or "Speaker X".

Identification works best when audio is clear, speakers do not overlap, and participants are in your contacts.

Language support​

Harmony detects the main language automatically. English, Spanish, French, and German are fully supported. In meetings with multiple languages, the primary language is transcribed most accurately.

Exporting transcripts​

Open the meeting and click Export (or the download icon). Choose PDF, DOCX, TXT, or JSON. You can control whether to include timestamps, speaker labels, and the summary.

Improving accuracy​

Good audio gives better transcripts. Use a headset or external mic and add key terms to Custom Vocabulary. Editing early on teaches Harmony your terminology and improves later meetings.

Troubleshooting​

Poor or wrong words: Check that the recording has clear audio. Add terms to Custom Vocabulary and fix errors in the transcript.

Wrong or missing speaker: Identification depends on clear audio and known attendees. Correcting speaker assignments is not yet available.

Missing segments: Check for gaps in the recording or very quiet sections.

Privacy​

Transcripts are encrypted at rest and in transit. Access follows your workspace permissions.

Harmony does not use your content to train AI models.