Voice
Overview
The Voice setting under Orchestrator allows administrators to configure how the Leena AI assistant handles voice-based interactions — both how it listens and how it speaks. There are two sub-settings:
- Transcription Settings — controls how user speech is captured and processed.
- Select Voice — controls the voice persona used by the assistant when responding.
How to Access
- Log in to the Leena AI Admin Dashboard
- Navigate to Settings.
- From the left sidebar, navigate to Orchestrator.
- Click on Voice.
- The Voice settings page loads with two collapsible sections — Transcription Settings and Select Voice.
Access: Only users with Admin access to the dashboard can view and edit Voice settings.
Transcription Settings
Transcription Settings tune the quality and behaviour of speech-to-text processing during voice interactions.
| Settings | Description |
|---|---|
| Denoising Mode | Controls background noise suppression. Remove noise filters ambient environmental noise. Remove noise + background speech additionally suppresses other voices audible in the environment — recommended for open office or shared space deployments. |
| Vocabulary Specialisation | Sets the domain model for transcription accuracy. General is the default and works for most enterprise use cases. Medical is optimised for medical terminology. |
| Boosted Keywords | A customisable list of keywords that expand the transcription model's vocabulary. Use this to improve recognition of organisation-specific terms, product codes, acronyms, or jargon that the default model may not handle reliably. |
| Additional Instructions | Free-text instructions to guide voice orchestrator behaviour — for example, how to handle filler words, specific input formats (e.g., reading out alphanumeric codes digit by digit), small talk handling, or response style. |
| Small Talk Frequency (sec) | How frequently (in seconds) the assistant engages in small talk during an active voice call session. Accepts a value between 3 and 8 seconds. |
Note: Boosted keywords are especially useful when users are likely to speak product codes, employee IDs, or internal terminology. Enter them as a comma-separated list to improve transcription recognition accuracy.
Select Voice
The Select Voice section allows administrators to choose the voice persona the assistant uses for spoken responses. 64 voices are available across different accents, age styles, and genders, letting you personalise the assistant to match your organisation's tone and user base.
| Column | Description |
|---|---|
| ▶ (Play) | Preview button — click to hear a sample of the voice before selecting. |
| Avatar | Visual avatar representing the voice persona. |
| Voice | Name of the voice persona (e.g., Cimo, Adrian, Amy (UK), Amritanshu (en-IN)). |
| Traits | Tags showing the voice's accent (e.g., American, British, Indian, Mexican) and age style (e.g., Young, Middle Aged). |
| Action | Select voice button to activate the persona for your bot. The currently active voice shows a ✓ Voice selected indicator. |
Note: Always use the preview button to listen before selecting. Accent and tone significantly affect user experience — choose a voice that feels natural for the language and region of your primary user base.
Updated 1 day ago