Meeting Transcription
Harmony turns meeting audio into a full, timestamped, speaker-labelled transcript. You can read it, search it, jump from any word to that moment in the recording, and export it via the Public API.
What you get
Every conversation processed from an audio or video recording includes a full, speaker-labelled transcript. Open the meeting and click the Transcript tab.
The transcript is displayed as a series of turns — each turn is one continuous block of speech by a single speaker. Each turn shows:
- Speaker label — who is speaking (real name when known, otherwise Speaker A, Speaker B, …)
- Timestamp — the time in the recording when this turn begins
- Text — the spoken words
The transcript view also shows a Stats section with each speaker's word count and percentage of the conversation.
Speaker labels
Speakers are assigned codes automatically based on detection order — single letters first (A, B, C, … up to Z), then two-character codes (AA, AB, …) for meetings with more than 26 participants.
When Harmony can match a speaker to a known Harmony user or a linked contact, the participant's real name appears instead of the generic code. Unmatched speakers fall back to the label "User", not "Unknown".
Speaker assignment works best when the audio is clear, speakers do not overlap, and known participants are in your contacts.
Reading the transcript
If the conversation has a recording, the transcript and the audio player are synchronised:
- Click any word to jump to that exact moment in the recording.
- As the recording plays, the transcript auto-scrolls to follow along (toggle from the Auto-scroll switch in the toolbar).
- The current speaking turn is highlighted as the audio progresses.
Use the Previous / Next buttons in the transcript toolbar to step between turns.
Search inside the transcript
Use your browser's built-in Find (Ctrl+F / Cmd+F) to search within the transcript view. From the Meetings list, the search bar searches across conversation titles and content — see Finding and searching meetings.
Closed captions
Toggle Closed Captions in the player controls (the Subtitles icon) to overlay live captions on the recording. They update word by word as the recording plays and can be toggled on or off at any time.
Editing the transcript
Transcript text is read-only in the Harmony app today. There is no inline click-to-edit feature, and Harmony does not "learn from corrections" you make in the UI.
What you can do:
- Match a speaker to a contact or workspace user from the conversation detail view. This updates which participant is attributed to those turns.
- Reprocess the conversation from the action menu (⋯ → Restart from Transcription) if the transcript is so far off that it is unusable. Reprocessing re-runs transcription against the original recording.
Transcript line-level commenting is planned as a future feature and is not yet available.
Translation
Translation of an existing transcript into another language is currently in development. When you click Translate in the transcript toolbar, Harmony shows a "Translation Coming Soon" notice.
Exporting the transcript
To export or access the transcript from Harmony, you have a couple of primary options, each serving different needs.
First, you can use the Share PDF function, which is accessible from the Share menu located next to the conversation title in the meeting view. This feature generates a downloadable PDF report that includes key insights produced by Harmony for the conversation, as well as relevant conversation metadata like the meeting title, date, participants, and any configured AI-driven summaries or action items. However, it's important to note that this PDF report intentionally does not contain the full transcript text for the conversation. If you are looking to share a quick summary of the meeting, communicate key takeaways, or archive high-level insights for external review, the PDF export is the right choice. For more detailed information about how to share meetings in various ways, see the Sharing meetings documentation.
If your requirement is to obtain or work with the entire transcript itself—for example, if you need the raw conversation text for record-keeping, further analysis, or integration with another tool—Harmony provides access through its Public API. You can make a request to GET /v1/conversations/{id}/transcript to retrieve the complete transcript as produced in the current Transcription Report. The result will be returned in JSON format, making it suitable for programmers, data analysts, or anyone needing a structured export for downstream processing or archiving outside of the Harmony app. This method is recommended when full transcript access is essential rather than a summary or high-level report.
As you interact with transcriptions in Harmony, you may encounter several distinct states or error messages, each indicating the progress or outcome of the transcription process. When you open a conversation that is still being processed, you might see a notification stating, "No transcription available yet. The recording is being processed." This means that Harmony is still generating the transcript, and you simply need to wait—once processing is finished, the transcript will automatically appear on the page without any need for manual refresh.
If the transcription process fails, the application will show an error message such as "Transcription failed. Please try again." In this case, you have the option to reinitiate the process from the action menu, by selecting the "Restart from Transcription" entry (accessed via the ⋯ menu).
Occasionally, you may encounter a loading error, signaled by the message "Failed to load transcription." When this happens, refreshing the page usually resolves the issue; however, if the error continues, it's best to reach out to Harmony support for assistance, as there could be an underlying technical problem.
Finally, there are situations where no transcript is available at all. In this scenario, you may see the notice "No transcript available," which generally means the conversation originated from a capture method that doesn't produce a transcript—such as certain types of uploads or notes—or that the meeting has not started processing yet.
Understanding these statuses and messages will help you diagnose what is happening with your conversation’s transcription and guide you on the right next steps to take, whether that's waiting, retrying, or contacting support.
Improving accuracy
Transcript accuracy depends on the source audio. To improve quality:
- Use a headset or external microphone.
- Reduce background noise.
- Avoid talking over one another.
- Have each speaker say their name once at the start of the call so participants can be matched to contacts.
There is no shipped custom-vocabulary feature today; if specialist terminology comes through phonetically, the most reliable workaround is clear pronunciation. See Transcription quality for the full list of factors and remediation steps.
Privacy
Transcript access follows your workspace's role and permission model — see Roles and Permissions. For details on how AI providers process meeting content, see Models and training data.