Text-to-Speech

Google TTS models are called via /v1/audio/speech. The response is raw audio bytes.

Example

curl https://llmtr.com/v1/audio/speech \
  -H "Authorization: Bearer llmtr-your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/gemini-2.5-flash-preview-tts",
    "input": "Read today\'s summary in a calm, clear voice.",
    "voice": "Kore"
  }' --output speech.wav

Parameters

| Field | Required | Description | |---|---|---| | model | yes | Canonical TTS model ID | | input | yes | Text to read (UTF-8) | | voice | yes | Voice character (model-specific list) | | speaker | no | Speaker name for multi-speaker output | | response_format | no | wav, mp3, opus, aac, flac | | speed | no | 0.25 - 4.0 |

voice is required

Gemini TTS models have no default voice. A request without voice is rejected with a 400 whose message names the missing field and gives an example value:

{
  "error": {
    "message": "\"voice\" is required for google/gemini-2.5-flash-preview-tts. Send it in the request body, for example: {\"model\": \"google/gemini-2.5-flash-preview-tts\", \"input\": \"Hello\", \"voice\": \"Kore\"}. Available voices include Kore, Puck, Charon, Aoede, Fenrir, Leda, Orus and Zephyr.",
    "type": "invalid_request"
  }
}

Voice characters

Example voices for Gemini TTS: Kore, Puck, Charon, Aoede, Fenrir, Leda, Orus, Zephyr. Full list is on the model detail page.

xAI TTS: language is required

xai/grok-voice-tts uses the same endpoint but requires language instead of voice. When it is missing the request is likewise rejected with a 400 and an example.

curl https://llmtr.com/v1/audio/speech \
  -H "Authorization: Bearer llmtr-your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "xai/grok-voice-tts",
    "input": "Read today\'s summary in a calm, clear voice.",
    "language": "en"
  }' --output speech.wav

language is an ISO 639-1 code (tr, en). voice is optional; the default voice is used when it is omitted.

Pricing

Billed per character or per second depending on the model. See the model card's native unit.