Managing speakers in a transcript

When Klang transcribes a recording, it separates the conversation into speakers. You can name each speaker, correct lines that went to the wrong person, and rerun speaker identification if the detection went wrong.

How naming works depends on whether People is turned on for your workspace. With People on, each speaker is linked to a person, and the name follows that person into every conversation. With People off, you name speakers separately in each transcript. Admins can change this, see Turning People and Speaker ID on or off.

How speakers appear in a transcript

Open a transcription from your analysis. At the top of the transcript you will see colored chips for each detected speaker. Below that, each segment shows the speaker name, a timestamp, and the transcribed text.

The colored chips act as quick filters. Click a chip to scroll to that speaker’s first segment.

Name a speaker

1. In the transcript, click the speaker name next to any line. The Who’s speaking? menu opens.
2. Optional: To hear the voice, click Listen to the voice
3. Choose the person from the list, or type their name in Search people
4. If the person is not in the list, click Add a new person and enter their name

The name updates on every line for that voice. The list starts with the people most likely to be speaking, such as others in the same conversation and people Klang recognizes.

If someone with the same name already exists, Klang asks you to listen to the voices. Choose the existing person, or create a new person with the same name.

Correct a line with the wrong speaker

1. Click the speaker name on the line
2. Choose where the change applies:
   • Only this row: Changes only the selected line.
   • This voice throughout the conversation: Changes every line for this voice.
3. Choose the correct person

For lines labeled Unknown, the second option is All rows labeled Unknown instead. It changes every line that is still marked Unknown.

To rename a person in every conversation at once, click the Show person / rename everywhere button in the menu. For more, see Managing people in your workspace.

When People is turned off

With People off, the speaker menu lists the speakers of the current transcript.

1. Click the speaker name next to any line

2. To rename a speaker, click the edit icon (pencil), type the new name, and click the send icon
3. To move the line to another speaker, click that speaker’s name
4. To add a speaker, type the name in Add speaker and click the send icon

If the same name shows up twice with a suffix like (0) and (1), speaker identification found more voices than there are people. If it is the same person, rename both to the same name without the suffix, and Klang merges them. If they are two different people, give each a unique name.

Rerun speaker identification

If the automatic speaker detection made too many errors, you can rerun it with different settings instead of fixing each segment one by one.

1. In the transcript toolbar, click the Settings icon (gear)

2. Click Rerun speaker identification
3. Choose the number of speakers, or leave it on Automatic
4. Click Run

Klang reprocesses the audio and merges the new speaker assignments into your existing transcript. Names you set before the rerun are kept where possible.

Note: Rerunning speaker identification replaces the current speaker assignments. If you have already made many manual corrections, consider fixing the remaining errors by hand instead.

Tips for better speaker identification

• Record video meetings with the Klang bot. When Klang joins a Google Meet, Microsoft Teams, or Zoom call, it picks up participant names directly from the meeting.
• Set the number of speakers when uploading if you know how many people are in the recording.
• Use clear audio. External microphones or separate audio channels give better results than a single laptop microphone.
• Tell Klang which speaker is you. See Recognizing speakers across conversations.