更新
What is Grok Text to Speech?
Grok Text to Speech is xAI's voice model, available on UnificAlly as xai/grok-tts alongside the Grok language and Grok Imagine models on the same key and balance. Send text and a voice ID, and the API returns a hosted audio file in mp3, wav, or raw pcm.
It covers twenty languages with automatic detection, a speed range of 0.7 to 1.5, and a catalog of preset voices with distinct styles, from warm assistant voices to dramatic narration. Billing is per character, so the cost of any request is known before it runs.
Key features of Grok Text to Speech
Preset voices with character
Each Grok voice comes with a style, from warm and friendly to authoritative or dramatic. Pick by ID; ara is the default. The same IDs work for spoken lines in Grok Imagine 1.5 Video.
Twenty languages, auto-detected
Leave language on auto and the model reads the text in its own language, or pin one with a BCP-47 code such as pt-BR when the text is ambiguous.
Speed from 0.7 to 1.5
Slow a voice down for accessibility and learning content or speed it up for notifications and summaries, without changing pitch or voice.
mp3, wav, or pcm
mp3 for delivery, wav for editing, raw pcm for pipelines that stream or post-process audio themselves.
Up to 60,000 characters per request
Long articles and scripts go through in one call. The response reports the number of characters billed.
What Grok Text to Speech is best for
Voiceovers for generated video
Narrate Grok Imagine or any other clip with a consistent voice, then mix the file in your editor.
Product and support voices
Order updates, confirmations, and IVR prompts in a warm assistant voice, in the language of the message.
Articles and newsletters as audio
Turn long text into listenable audio in one request, with per-character pricing that scales with length.
Multilingual content
One request per language with the same voice ID keeps a brand voice consistent across markets.
Grok Text to Speech limitations
Voices are preset: there is no voice cloning or custom voice upload on this route. Speed runs from 0.7 to 1.5 only, and there are no fine-grained prosody or emotion controls beyond punctuation and phrasing in the text. Streaming playback is not part of the task API; the result is a complete file URL once the task finishes.
Grok Text to Speech API examples
Call Grok Text to Speech from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/audio/xai/grok-tts.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "xai/grok-tts",
"input": {
"text": "Welcome back. Your order has shipped and should arrive on Thursday.",
"voice_id": "ara",
"language": "en",
"speed": 1.0,
"format": "mp3"
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request; the completed task carries the audio url and the number of characters billed.
FAQ
Grok Text to Speechに関するよくある質問
It is xAI's Grok voice model served through UnificAlly as xai/grok-tts. Send text, pick a preset voice, and get back an audio file. Language is detected automatically or set with a BCP-47 code, speed runs from 0.7 to 1.5, and the output is mp3, wav, or raw pcm.
A catalog of Grok preset voices, each with a short style description; ara is the default. The playground lists them, and the same voice IDs work in the video routes that accept preset voices.
Twenty, including English, Spanish, Portuguese, French, German, Italian, Japanese, Korean, and Chinese. Leave language on auto to detect it from the text, or pass a code such as en, es, or pt-BR to force one.
Per character of the text you submit, so the cost of a request is known before it runs. The playground shows the price for your text before you generate.
Up to 60,000 characters per request. For a long script, split on paragraph boundaries and join the files, or send one request per section.
mp3 by default, or wav and raw pcm on request through the format field. The result is a hosted file URL plus the character count that was billed.
