By registering members’ voice data (voice prints) in advance, AI can identify speakers during transcription and automatically separate them. This is a useful feature for internal company meetings where the same members regularly participate.
Flow for Using Automatic Speaker Separation with Voice Data
Step1│ Turn on Automatic Speaker Separation in Settings
Step2│ Register Voice Data
Step3│ Execute Automatic Speaker Separation
Step1│ Turn on Automatic Speaker Separation in Settings (Administrator Setting)
Click “Speaker Separation” from the global menu “Settings.”
Turn on “Automatically add speakers as meeting participants” and “Automatic Speaker Separation” (move the slider to the right to make it orange 🟠)
Step2│ Register Voice Data
①How to Register Your Own Voice Data
1. Click the circular icon in the upper right corner of the screen, then select “Profile” from the displayed menu.
2. Click “Record Now.”
3. Review the registration instructions and click “Start Recording.”
4. A voice print registration note will open. Read the text shown on the left side of the screen. After speaking for 50-60 seconds, click the “End” button in the orange frame at the top right.
Note that recording ends automatically after 1 minute and the screen will close.
5. The “Voice Print Registration Note” you just read will appear in “Registered Audio.”
*If it doesn’t appear, please reload the page.
*This may take some time to process.
6. Click “Edit” on the “Voice Print Registration Note” to open the note, then select the “Transcription” tab to check the status of your voice data registration.
②How to Register Members’ Voice Data
Register voice data for all members you want to use for speaker separation from the Members page.
Voice print candidates are taken from sections where speakers have been set in previous notes (transcriptions/minutes).
For optimal accuracy, we recommend registering about 5 entries per person.
1. From the global menu “Members,” click on the tile of the person whose voice data you want to register to open their “Profile.”
Click on “AI Learning Voice Data” at the bottom of the “Profile” section.
2. A list of “Notes Containing Your Speech” (voice print candidates) will be displayed. Click “Go to Note.”
If “Notes Containing Your Speech” is not displayed: The person has not been set as a speaker in notes created so far. Please register them as a speaker in several notes.
3. A note will open in a new window with a “Speaker Separation” area on the right.
Click “Register Voice Data for Speaker Identification” at the bottom.
4. Under “Voice Data Registration,” speech segments will be displayed.
Select sections where the member’s voice is clearly recorded, click “Register,” and mark as “Registered.”
(Click the play button ▶ to check the audio)
💡 If the name of the person you want to register doesn’t appear in “Voice Data Registration,” check whether there’s a ☑︎ next to the speaker information on the transcription screen.
Checks can be added when in “Edit” mode. Please verify that all utterances belong to that person before adding the check.
5. Return to the page of the member whose voice data you registered and reload. The section you just registered will appear in “Registered Audio.”
(Click “Edit” to reopen the note and modify voice data registration if needed)
Step3│ Execute Automatic Speaker Separation
1. Upload the audio/video file you want to transcribe. After transcription is complete, participants will be automatically added and speaker separation will be performed.
*Voices of people without registered voice data will be recognized as “Unnamed Speaker.”
2. If there are participants whose voice prints are not registered, manually set the speaker for their speech segments and execute speaker separation again.
This completes the procedure for registering voice data (voice prints) and performing automatic speaker separation.
For information on how to perform speaker separation without registering voice data, please refer to the section below.