Voices

Every agent speaks in one voice, chosen by its voice.voice_id. This page is the honest answer to “where do the voice IDs come from?”.

Finding a voice_id

v2 has no voice-list endpoint — this is the current path

There is no GET /v2/voices today. The voice catalogue is shared with v1, and that is where you list it. The voice_id you use on a v2 agent is the exact slug this endpoint returns.

GET /v1/admin/voice-gallery

curl https://api.voice-agents.miraiminds.co/v1/admin/voice-gallery \
  -H "x-public-key: pk_…" \
  -H "x-private-key: sk_…"

200 OK

{
  "status_code": 200,
  "message": "Voices fetched successfully.",
  "data": [
    {
      "_id": "6690a1b2c3d4e5f600000301",
      "slug": "priya",
      "sampleAudio": "https://cdn.miraiminds.co/voices/priya.wav"
    }
  ]
}

The catalogue uses v1 admin auth (x-public-key + x-private-key), not your v2 sk_live_ secret key. If you do not have v1 admin credentials, ask ops — email sneh@miraiminds.co — and we will send the current list. Full reference: v1 Voice gallery.

Listen to sampleAudio before you pick. Voice choice moves answer-through rates more than prompt wording does.

Commonly used slugs include priya, neha and sagar. Treat the gallery, not this page, as the source of truth — the catalogue can grow, and a slug that is not in the gallery is rejected at agent-create time with 400 invalid_request (unknown voice_id).

Setting a voice on an agent

Put the slug in voice.voice_id when you create or update an agent:

{
  "name": "Order confirmations",
  "system_prompt": "…",
  "voice": { "voice_id": "priya", "language": "hi-IN" },
  "language": "hi-IN"
}

Field	Type	Description
voice_id	string	A voice slug from the gallery, e.g. priya. Required. Unknown slug ⇒ 400.
language	string	Rendering language for the voice. Omit to inherit the agent’s top-level language. Accepted and stored, not yet honoured — see below.

Change a voice later with PATCH /v2/agents/{id}:

curl -X PATCH https://api.voice.miraiminds.co/v2/agents/agt_… \
  -H "Authorization: Bearer sk_live_…" \
  -H "Content-Type: application/json" \
  -d '{ "voice": { "voice_id": "neha", "language": "hi-IN" } }'

See the Agents reference for the full agent shape.

The language caveat

voice.language is recorded, not yet applied

voice.language is validated, stored, echoed back and carried with the call payload — but the voice today renders in the agent’s top-level language. Set both to the same value and you get what you expect; set them differently and the top-level one wins until per-voice rendering ships.

The practical rule: keep voice.language and the agent’s language identical.

Premium / vendor voices

A wider set of voices (a premium TTS option with automatic fallback) is a t3 and t5 feature. Those tiers are published but not live — see Billing & tiers. On t1 today, choose from the shared gallery above.