> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mavioapp.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Speaker Profiles

> Create voice profiles for automatic speaker identification and cross-meeting tracking.

Speaker profiles let Mavio recognize who is speaking in every meeting. By enrolling a voice sample, Mavio builds a unique voice fingerprint for each person. Once a profile exists, Mavio automatically labels transcript segments with the correct speaker name across all future meetings.

## How speaker identification works

Mavio uses voice fingerprinting to distinguish between speakers. The system analyzes vocal characteristics like pitch, cadence, and tone — not the words spoken — to create a unique acoustic signature for each person.

<Steps>
  <Step title="Voice enrollment">
    A speaker records a short voice snippet (15-30 seconds) or Mavio extracts a sample from an existing meeting recording.
  </Step>

  <Step title="Fingerprint generation">
    Mavio AI processes the audio to create a compact voice fingerprint that captures the speaker's unique vocal characteristics.
  </Step>

  <Step title="Real-time matching">
    During future meetings, Mavio compares incoming audio against enrolled fingerprints to label each speaker in the transcript.
  </Step>

  <Step title="Continuous improvement">
    As more meetings are processed, Mavio refines each speaker's fingerprint for increasingly accurate identification.
  </Step>
</Steps>

## Creating a speaker profile

### From a meeting recording

1. Open any meeting with the speaker present.
2. Navigate to the **Transcript** tab.
3. Click on a correctly labeled speaker segment.
4. Select **Create Speaker Profile** from the context menu.
5. Enter the speaker's name, email, and optionally their role or organization.
6. Mavio extracts a voice sample from the selected segment.

### From a dedicated recording

1. Go to **Settings > Speaker Profiles > Add Profile**.
2. Enter the speaker's details (name, email, role).
3. Click **Record Voice Sample** and have the speaker talk naturally for 15-30 seconds.
4. Mavio processes the recording and creates the profile.

<Tip>
  For the best accuracy, enroll voice samples from high-quality audio. A quiet environment with a good microphone produces the most reliable fingerprints.
</Tip>

## Managing profiles

### Profile information

Each speaker profile contains:

| Field              | Description                                                  |
| ------------------ | ------------------------------------------------------------ |
| **Name**           | Display name used in transcripts                             |
| **Email**          | Linked email address for matching with calendar participants |
| **Role**           | Optional job title or role for context                       |
| **Organization**   | Optional company or team name                                |
| **Voice snippets** | One or more enrolled voice samples                           |
| **Meetings**       | List of meetings where this speaker was identified           |

### Editing a profile

Open **Settings > Speaker Profiles**, select a profile, and update any field. Changes to the display name are applied retroactively to all past transcripts.

### Merging duplicate profiles

If Mavio creates separate profiles for the same person (common when someone joins from different devices):

1. Open **Settings > Speaker Profiles**.
2. Select one profile and click **Merge**.
3. Select the duplicate profile to merge into the first.
4. Voice snippets from both profiles are combined. Past transcripts are updated.

## Voice snippets

Each profile can have multiple voice snippets to improve accuracy across different conditions:

* **Quiet environment** — standard voice sample
* **Noisy environment** — sample with background noise
* **Phone audio** — sample from a phone or low-quality microphone
* **Different energy levels** — samples from calm discussion and animated debate

Add more snippets from **Settings > Speaker Profiles > \[Profile] > Add Voice Snippet**.

## Cross-meeting speaker tracking

Once a profile is created, Mavio tracks that speaker across all meetings:

* **Meeting history** — see every meeting a speaker has participated in
* **Talk time analytics** — total and average speaking time per meeting
* **Topic associations** — what subjects this speaker most frequently discusses
* **Interaction patterns** — who they speak with most often

Access these analytics from the speaker profile page or from the **Speakers** tab in any meeting.

## Speaker analytics

| Metric                      | Description                                                |
| --------------------------- | ---------------------------------------------------------- |
| **Total talk time**         | Cumulative speaking time across all meetings               |
| **Average talk ratio**      | Percentage of meeting time this speaker typically occupies |
| **Meeting count**           | Number of meetings the speaker has appeared in             |
| **Most common co-speakers** | Other speakers frequently in the same meetings             |
| **Topic frequency**         | Most discussed topics by this speaker                      |

## Advanced settings

<AccordionGroup>
  <Accordion title="Adjusting identification sensitivity">
    In **Settings > AI > Speaker Identification**, adjust the confidence threshold. A higher threshold produces fewer false matches but may leave some segments as "Unknown Speaker". A lower threshold catches more segments but may occasionally misidentify speakers.
  </Accordion>

  <Accordion title="Handling unknown speakers">
    When Mavio encounters a voice that does not match any profile, it labels the speaker as "Speaker 1", "Speaker 2", etc. After the meeting, you can assign these labels to existing profiles or create new ones.
  </Accordion>

  <Accordion title="Privacy considerations">
    Voice fingerprints are stored securely and encrypted at rest. They are never shared outside your workspace. Speakers can request deletion of their voice profile from **Settings > Speaker Profiles > \[Profile] > Delete**.
  </Accordion>

  <Accordion title="Improving accuracy for similar voices">
    If Mavio confuses two speakers who sound similar, enroll additional voice snippets for both. Snippets from the same meeting where both speakers are present are especially helpful for training the model to distinguish them.
  </Accordion>
</AccordionGroup>

<Note>
  Speaker profile enrollment consumes 3 AI credits per voice snippet processed. See [AI Credits](/documentation/ai/ai-credits) for details.
</Note>