Voice settings

1
2
3
4
5
6

{
  "title": "Voice settings",
  "audience": "power_user",
  "summary": "Choose your AI's voice, set style and speed, and configure the tenant voice allowlist.",
  "sort_order": 10
}

Voice settings#

ScaiWave's voice features (read-aloud, dictation, voice calls) use ScaiSpeak for text-to-speech and ScaiEcho for speech-to-text. You can choose which voice the AI speaks with and, if you're an admin, which voices are available to your team.

Choosing your default voice#

1. Settings → scroll to the Voice (read aloud) section.
2. Pick a voice from the dropdown. Each entry shows the voice name
   • primary language.
3. Your choice is saved immediately. It applies to:
   • The "read aloud" button on chat messages
   • Voice calls with the AI

If you don't pick one, the tenant default applies. If no tenant default is set, ScaiWave picks the first available voice.

Resolution chain#

When a voice is needed, ScaiWave resolves it in this order:

1. Your explicit choice (if you picked one in Settings)
2. Tenant default (set by your admin)
3. First available voice (from whatever voices your tenant can see)

If you pick a voice and the admin later removes it from the allowlist, your next request falls through to the tenant default without error.

Voice style and speech speed#

Beyond choosing which voice the AI uses, you can control how it speaks during voice calls.

1. Settings → Voice → scroll to Voice style.
2. Pick one of the built-in presets:
   • Warm & professional — calm, authoritative, clear
   • Cheerful & energetic — upbeat, friendly, animated
   • Calm & soothing — relaxed, gentle, measured pace
   • Broadcaster — polished, confident, news-anchor delivery
3. Or choose Custom and type your own style instructions (e.g. "Speak like an enthusiastic museum guide").

Below the style picker, the Speech speed slider lets you adjust from 0.5x (half speed — useful for language learners or complex material) to 2.0x (double speed — for quick recaps or when you're in a hurry). The default is 1.0x.

Both settings are per-user and stored alongside your voice selection. They take effect on the next voice call — changing them mid-call doesn't alter the current session.

These settings apply only to live voice calls. The "read aloud" button on chat messages uses its own fixed style (see Read messages aloud).

Tenant admin: voice allowlist#

Admins can restrict which voices end-users see:

1. Admin panel → Voice tab.
2. You'll see the full catalog of voices available to your tenant (Global + Tenant-scoped voices from ScaiSpeak).
3. Check / uncheck individual voices. Only checked voices are visible to non-admin users.
4. "Show all voices" clears the allowlist entirely — every voice visible.
5. Set a tenant default voice so users who haven't picked one still get a consistent experience.

The allowlist is a UX guardrail, not a security boundary — it controls what appears in the voice picker UI.

Voice permissions#

Voice features (TTS + STT) require ScaiSpeak/ScaiEcho module permissions on the ScaiGrid side. If you see "Voice features aren't enabled for your account," your platform admin needs to grant scaispeak:voice.read + scaispeak:synthesize + scaiecho:transcribe to your user or role.

Where to go next#

• Talk to your AI — voice conversations
• Dictate messages and notes — mic button
• API: Voice endpoints — REST reference