Grok Voice Agent API, Grok Voice TTS API, Grok Voice STT API: Pricing, Documentation

Authentication

All requests require an API key passed via header.

Webhook (Optional)

Instead of polling, you can receive a Webhook callback when the request reaches a terminal state. Provide a Webhook URL via header on the submit request.

Example: enable Webhook

curl -X POST 'https://gateway.pixazo.ai/grok-voice/v1/speech-to-speech' \
  -H 'Content-Type: application/json' \
  -H 'Ocp-Apim-Subscription-Key: YOUR_SUBSCRIPTION_KEY' \
  -H 'X-Webhook-URL: https://your-server.com/webhook' \
  --data-raw '{"audio_url": "https://pub-582b7213209642b9b995c96c95a30381.r2.dev/doc-assets/audio/transcribe-sample.wav", "prompt": "You are a friendly assistant. Answer briefly and concretely.", "voice": "eve"}'

Callback Payload (success)

{
  "request_id": "grok-voice_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "COMPLETED",
  "model_id": "grok-voice",
  "output": {
    "media_url": [
      "https://pub-582b7213209642b9b995c96c95a30381.r2.dev/v1/{request_id}/output.wav"
    ],
    "media_type": "audio/wav"
  },
  "created_at": "2026-09-21T09:14:16.102Z",
  "completed_at": "2026-09-21T09:14:31.870Z"
}

Failure callback shape

{
  "request_id": "grok-voice_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "ERROR",
  "model_id": "grok-voice",
  "error": "Description of the failure"
}

Delivery semantics

• terminal mode: one Webhook callback when the request is COMPLETED or ERROR.
• sync mode: a Webhook callback on each status change.
• Callbacks are idempotent on request_id — de-duplicate on it.
• Respond 200 within a few seconds; the Webhook endpoint must be HTTPS.

Request Parameters

Audio requirements

• At most 10 minutes of audio and at most 50 MB per request. Longer recordings fail once the file is read, reported as status: "ERROR".
• Most common formats are accepted (WAV, MP3, M4A, OGG, FLAC and similar). The audio is converted to 16-bit 24 kHz mono PCM before it reaches the model, so you do not need to convert it yourself.
• audio_url must be an https URL and publicly reachable.
• Cost scales with the length of the reply the agent generates, not the recording you send; the current rate is in the pricing panel on this page.

Fields this endpoint does not accept

The provider's agent tools (tools.web_search, tools.x_search, tools.mcp_servers) are not available on this endpoint: a body containing tools is rejected with a synchronous 400. Likewise any field that is not listed in the table above, such as duration or seconds, is rejected rather than ignored.

Voices

Pass one of these 28 names as voice, exactly as written (lowercase). Omit the field to use Grok's default voice.

carina, zagan, helix, orion, luna, iris, altair, zenith, perseus, helios, lux, kepler, rigel, cosmo, celeste, ursa, sirius, lumen, castor, naksh, atlas, aurora, liora, ara, eve, leo, rex, sal

Request Headers

Response Handling

Only what can be judged from the request body itself is rejected synchronously. Everything that needs the audio to be fetched and read is reported through the status endpoint instead, as status: "ERROR". Failed requests are not billed — the hold is released in full.

Retrieving Results

Poll the status endpoint with the request_id from the submit response until status is COMPLETED (or FAILED/ERROR), then download the reply audio from output.media_url.

curl 'https://gateway.pixazo.ai/v2/requests/status/grok-voice_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx' \
  -H 'Ocp-Apim-Subscription-Key: YOUR_SUBSCRIPTION_KEY'

Completed response

{
  "request_id": "grok-voice_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "COMPLETED",
  "model_id": "grok-voice",
  "output": {
    "media_url": [
      "https://pub-582b7213209642b9b995c96c95a30381.r2.dev/v1/{request_id}/output.wav"
    ],
    "media_type": "audio/wav"
  },
  "created_at": "2026-09-21T09:14:16.102Z",
  "completed_at": "2026-09-21T09:14:31.870Z"
}

The reply is returned as audio only — a 24 kHz mono WAV file. A text transcript of the reply is not returned by this endpoint; if you need one, pass the WAV to a speech-to-text API.

Response Fields

Status Values & Flow

QUEUED → PROCESSING → COMPLETED (success) or FAILED/ERROR (failure).

How billing is measured

Billing is measured on the length of the reply audio the model generates, per second, rounded up to the next whole second — not on the length of the recording you send, and not on how long the request takes. A hold is placed when the request is submitted, because the reply length is not known until the model has answered; the hold is reduced to the real cost once the reply audio is ready, and released in full if the request fails. To keep replies short, say so in prompt.

The current rate is shown in the pricing panel on this page.

Authentication

All requests require an API key passed via header.

Webhook (Optional)

Add the X-Webhook-URL header to your submit request to receive a POST callback when the job completes — no polling required.

Using curl? These are HTTP request headers — pass each with -H, e.g. -H "X-Webhook-URL: https://your-server.com/webhook/callback". Do not paste them as bare lines, and end every line of a multi-line command with \.

Webhook Headers

Example: enable webhook

X-Webhook-URL: https://your-server.com/webhook/callback
X-Webhook-Mode: terminal

Callback Payload

Your endpoint receives a POST application/json with the same shape as the GET /v2/requests/status/{request_id} response. Example terminal callback (mode terminal):

{
  "request_id": "xai-text-to-speech_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "COMPLETED",
  "model_id": "xai-text-to-speech",
  "error": null,
  "output": {
    "media_url": [
      "https://pub-582b7213209642b9b995c96c95a30381.r2.dev/v1/xai-text-to-speech_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx/output.wav"
    ],
    "media_type": "audio/wav"
  },
  "created_at": "2026-05-22T13:17:32.110Z",
  "updated_at": "2026-05-22 13:19:23",
  "completed_at": "2026-05-22 13:19:23"
}

Failure callback shape

{
  "request_id": "xai-text-to-speech_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "ERROR",
  "model_id": "xai-text-to-speech",
  "error": "Description of the error",
  "output": null,
  "created_at": "...",
  "updated_at": "...",
  "completed_at": "..."
}

Delivery semantics

• terminal mode (default) — exactly one POST when the request reaches a terminal status. No callback during PROCESSING.
• sync mode — POST on every status poll (with delay capped at ~15s) plus a final POST at terminal status. Use when you want progress updates.
• Idempotency — use request_id as your idempotency key. Network retries can deliver the same callback more than once; your handler must tolerate duplicates.
• Response — respond 200 OK within a few seconds. The queue does not block on slow handlers, but persistent failures may stop further deliveries.
• HTTPS required — plain http:// URLs are rejected.

Request Parameters - Grok Voice Text to Speech generate request

Response

{
  "request_id": "xai-text-to-speech_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "QUEUED",
  "polling_url": "https://gateway.pixazo.ai/v2/requests/status/xai-text-to-speech_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"
}

Request Headers

Response Handling

Common status codes for Grok Voice Text to Speech generate request.

Error Responses

Queue system errors and model validation errors.

Queue System Errors

// 402 — Insufficient balance
{
  "error": "Insufficient Balance",
  "message": "Your wallet does not have enough balance."
}

// 400 — Model not found
{
  "error": "Model not found",
  "message": "Model 'xai-text-to-speech' not found or is disabled"
}

Error via Status/Webhook

{
  "request_id": "xai-text-to-speech_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "ERROR",
  "model_id": "xai-text-to-speech",
  "error": "Description of the error",
  "output": null
}

Retrieving Results

Poll the universal status endpoint to check progress and retrieve results.

Endpoint

GET https://gateway.pixazo.ai/v2/requests/status/{request_id}
Ocp-Apim-Subscription-Key: YOUR_API_KEY

curl -H "Ocp-Apim-Subscription-Key: YOUR_API_KEY" \
  "https://gateway.pixazo.ai/v2/requests/status/xai-text-to-speech_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx"

{
  "request_id": "xai-text-to-speech_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "COMPLETED",
  "model_id": "xai-text-to-speech",
  "error": null,
  "output": {
    "media_url": [
      "https://pub-582b7213209642b9b995c96c95a30381.r2.dev/v1/xai-text-to-speech_019dxxxx-xxxx/output.ext"
    ],
    "media_type": "application/octet-stream"
  },
  "created_at": "2026-03-31T10:00:00.000Z",
  "updated_at": "2026-03-31T10:00:15.000Z",
  "completed_at": "2026-03-31T10:00:15.000Z"
}

Response Fields

Status Values

Status Flow

QUEUED → PROCESSING → COMPLETED
                    → FAILED
                    → ERROR

Typical Workflow

1. Send a generate request to the API endpoint
2. Save the request_id from the response
3. Poll every 5-10 seconds: GET /v2/requests/status/{request_id}
4. When status is "COMPLETED", download from output.media_url

Tip: Use X-Webhook-URL header to get a callback instead of polling.

Authentication

All requests require an API key passed via header.

Webhook (Optional)

Instead of polling, you can receive a Webhook callback when the request reaches a terminal state. Provide a Webhook URL via header on the submit request.

Example: enable Webhook

curl -X POST 'https://gateway.pixazo.ai/grok-stt/v1/speech-to-text' \
  -H 'Content-Type: application/json' \
  -H 'Ocp-Apim-Subscription-Key: YOUR_SUBSCRIPTION_KEY' \
  -H 'X-Webhook-URL: https://your-server.com/webhook' \
  --data-raw '{"audio_url": "https://pub-582b7213209642b9b995c96c95a30381.r2.dev/doc-assets/audio/transcribe-sample.wav", "language": "en", "format": true, "diarize": true, "keyterm": ["Pixazo"]}'

Callback Payload (success)

{
  "request_id": "grok-stt_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "COMPLETED",
  "model_id": "grok-stt",
  "error": null,
  "output": {
    "media_url": [
      "https://pub-582b7213209642b9b995c96c95a30381.r2.dev/v1/{request_id}/output.json"
    ],
    "media_type": "application/json"
  },
  "created_at": "2026-11-10T08:22:53.327Z",
  "updated_at": "2026-11-10 08:24:27",
  "completed_at": "2026-11-10 08:24:26"
}

Failure callback shape

{
  "request_id": "grok-stt_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "ERROR",
  "model_id": "grok-stt",
  "error": "Description of the failure"
}

Delivery semantics

• terminal mode: one Webhook callback when the request is COMPLETED or ERROR.
• sync mode: a Webhook callback on each status change.
• Callbacks are idempotent on request_id — de-duplicate on it.
• Respond 200 within a few seconds; the Webhook endpoint must be HTTPS.

Request Parameters

Fields this endpoint does not accept

prompt is not supported by Grok Voice Transcribe 2.0 and is rejected with a synchronous 400; to steer how names and terms are recognised, use keyterm. Streaming and raw-audio fields (websocket, audio_format, sample_rate, channels, multichannel) are also rejected: send a complete recording file as audio_url.

Request Headers

Response Handling

Only what can be judged from the request body itself is rejected synchronously. Everything that needs the recording to be fetched and read is reported through the status endpoint instead, as status: "ERROR". Failed requests are not billed — the hold is released in full.

Retrieving Results

Poll the status endpoint with the request_id from the submit response until status is COMPLETED (or FAILED/ERROR), then download the transcript from output.media_url.

curl 'https://gateway.pixazo.ai/v2/requests/status/grok-stt_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx' \
  -H 'Ocp-Apim-Subscription-Key: YOUR_SUBSCRIPTION_KEY'

Completed response

{
  "request_id": "grok-stt_019dxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "status": "COMPLETED",
  "model_id": "grok-stt",
  "error": null,
  "output": {
    "media_url": [
      "https://pub-582b7213209642b9b995c96c95a30381.r2.dev/v1/{request_id}/output.json"
    ],
    "media_type": "application/json"
  },
  "created_at": "2026-11-10T08:22:53.327Z",
  "updated_at": "2026-11-10 08:24:27",
  "completed_at": "2026-11-10 08:24:26"
}

The transcript file (output.media_url)

A JSON document. This is the real transcript for the example request above, with diarize, format and keyterm on.

{
  "state": "Completed",
  "result": {
    "text": "The quick brown fox jumps over the lazy dog, Pixazo Gateway transcription test, 30 August 2026.",
    "language": "en",
    "duration": 7.649,
    "words": [
      {
        "text": "The",
        "start": 0.02,
        "end": 0.1,
        "speaker": 0
      },
      {
        "text": "quick",
        "start": 0.16,
        "end": 0.36,
        "speaker": 0
      },
      {
        "text": "brown",
        "start": 0.461,
        "end": 0.761,
        "speaker": 0
      },
      {
        "text": "fox",
        "start": 0.861,
        "end": 1.101,
        "speaker": 0
      },
      {
        "text": "jumps",
        "start": 1.201,
        "end": 1.541,
        "speaker": 0
      },
      {
        "text": "over",
        "start": 1.601,
        "end": 1.841,
        "speaker": 0
      },
      {
        "text": "the",
        "start": 1.901,
        "end": 2.001,
        "speaker": 0
      },
      {
        "text": "lazy",
        "start": 2.061,
        "end": 2.381,
        "speaker": 0
      },
      {
        "text": "dog,",
        "start": 2.441,
        "end": 2.781,
        "speaker": 0
      },
      {
        "text": "Pixazo",
        "start": 3.181,
        "end": 3.762,
        "speaker": 0
      },
      {
        "text": "Gateway",
        "start": 3.842,
        "end": 4.242,
        "speaker": 0
      },
      {
        "text": "transcription",
        "start": 4.322,
        "end": 4.982,
        "speaker": 0
      },
      {
        "text": "test,",
        "start": 5.042,
        "end": 5.462,
        "speaker": 0
      },
      {
        "text": "30",
        "start": 5.662,
        "end": 5.962,
        "speaker": 0
      },
      {
        "text": "August",
        "start": 6.022,
        "end": 6.307,
        "speaker": 0
      },
      {
        "text": "2026.",
        "start": 6.347,
        "end": 7.428,
        "speaker": 0
      }
    ]
  }
}

Response Fields

Transcript file fields

Status Values & Flow

QUEUED → PROCESSING → COMPLETED (success) or FAILED/ERROR (failure).

How billing is measured

Billing is measured on the length of the recording you send, per minute, rounded up to the next whole minute — a 15-second clip counts as one minute. The options (diarize, format, filler_words, keyterm) do not change the cost. A hold is placed when the request is submitted and reduced to the real cost once the transcript is ready; it is released in full if the request fails.

The current rate is shown in the pricing panel on this page.