👤 Speaker & Voiceprint Management

For multi-person meetings, we recommend building your speaker library first; it makes it much easier to distinguish "who is speaking" during transcription.

Last Updated: 2026-03-31 · Language: English

1. Adding a New Speaker

1. Go to "Speaker Management".
2. Click "Add". Speaker Name is required, Remark is optional.
3. Choose a recognition model for this speaker if needed. (Enabling Smart Mode will use this model for transcription)

2. Adding Sample Audio for a Speaker

1. Select a clear vocal file.
2. Set the start/end time in the cropping window and preview the audio.
3. Choose the language category (Chinese samples for Chinese, English samples for English, others based on linguistic family).
4. Click "Save". The system will automatically extract features and add them to the speaker profile.

3. Daily Maintenance

• A speaker can have multiple sample segments.
• You can switch between current samples and preview them directly.
• Modify names, remarks, or models at any time.
• When deleting a sample, the corresponding local file will also be deleted.

4. Why is no name displayed in offline results?

The voiceprint library mainly takes effect during Offline File Transcription. Please verify:

• Ensure the segmentation method is set to "Speaker Diarization".
• Ensure "Identity Tagging" is enabled.
• Ensure the voiceprint language selected during offline recognition matches the language chosen when adding the speaker sample.
• Lower the clustering threshold in offline settings.

5. Why is the name displayed incorrectly?

• Increase the recognition matching threshold.
• Re-add audio segments for the corresponding speaker to improve accuracy.