Custom Brand Voice · Cloning

Your brand’s own voice across every call and API.

Upload a clear 3–6 s reference clip (up to 11 s), automatically trim silence and align the transcript, and deploy the voice across all your agents and TTS routes with verified consent.

3 – 6 sIdeal clean reference recording length
11 sMax window with silence-boundary trimming
100%Explicit speaker consent required
IsolatedPrivate voices stay scoped to your tenant

From reference clip to live voice in 3 steps

01 · Upload & Consent

Clean clip + speaker consent

Upload or record a noise-free clip and attach speaker authorization before voice creation.

02 · Align

Silence trim & transcript verification

Trims leading/trailing silence and re-verifies the transcript so synthesis stays faithful and artifact-free.

03 · Deploy

Instant availability in Studio & API

Assign the cloned voice ID to any Voice Agent or call it directly via `/v1/audio/speech`.

Built with quality & safety guardrails

Strict tenant isolation

Your custom voices are never exposed publicly or shared across workspaces.

Reference-leak prevention

Automatic safeguards prevent reference prompt phrasing from leaking into generated replies.

Dialect continuity

Preserves speaker timbre across Jordanian, Saudi, and Lebanese agent prompts.

Warm cache ready

Pre-warms reference acoustic state so live calls reply without cold-start delay.

Common questions

How long should the reference recording be?

3 to 6 seconds of clean, noise-free speech gives the best clarity (trimmed automatically up to 11 s max).

Are custom voices ever visible to other customers?

No. Custom clones are strictly private to your tenant and never appear in the public voice library.