Grok Text-to-Speech

Grok Text-to-Speech: Text to Audio Model

What is Grok Text-to-Speech?

Grok Text-to-Speech is xAI's voice synthesis model that turns written text into natural, expressive spoken audio with a single API call. It is built on the same audio stack that powers Grok Voice in the X apps, Tesla in-car assistants, and Starlink support, then packaged as a standalone text-to-speech endpoint for developers. Send it a string of text and it returns ready-to-play audio in seconds, with control over voice, language, delivery, and output format. The model targets production voice work: voice agents, read-aloud features, podcasts, IVR systems, and accessibility tools.

Key Features

• •Five distinct voices (ara, eve, leo, rex, sal), each with its own personality from upbeat to authoritative.
• •20 languages with BCP-47 codes plus automatic language detection.
• •Inline speech tags like [pause] and [laugh] and wrapping tags like <whisper> for fine-grained delivery.
• •Multiple output codecs: mp3, wav, pcm, mulaw, and alaw for web, post-production, and telephony.
• •Up to 15,000 characters per request and an adjustable speed multiplier.

Best Use Cases

Reach for Grok TTS when you need believable narration at scale. It fits voice assistants and conversational agents, e-learning and audiobook narration, podcast intros, product demos, news read-aloud, and IVR or call-center prompts. Telephony codecs (mulaw, alaw at 8 kHz) drop straight into phone systems, while high-bit-rate mp3 and wav suit web players and post-production. In testing, the default eve voice produced clean, broadcast-quality English narration with natural pacing and no audible artifacts.

Prompt Tips and Output Quality

Write the way you want it spoken: commas, periods, and question marks guide intonation, and exclamation marks add energy. Drop inline tags where an expression naturally occurs, and wrap full phrases with delivery tags rather than single words. Keep requests under 15,000 characters, breaking long scripts into paragraphs for consistent pacing. Pick eve for upbeat demos, ara for warm support, rex for business, leo for instruction, and sal for balanced narration.

How to Use Grok Text-to-Speech

Grok Text-to-Speech converts written text into natural spoken audio in one call. Set the text you want spoken, then tune voice_id, language, codec, and speed to match your use case.

Choosing a voice

Each voice has a distinct personality. Use eve for energetic demos and announcements, ara for warm conversational support, rex for confident business narration, leo for authoritative instruction, and sal for smooth, balanced delivery. Voice IDs are case-insensitive.

Setting language

Pass a BCP-47 code such as en, es-ES, fr, de, or hi to lock the language for consistent results, or use auto to let the model detect it. The model covers 20 languages, so match the code to your script.

Writing expressive text

Punctuation drives prosody, so write clean sentences with commas, periods, and question marks. Add inline tags such as [pause], [long-pause], and [laugh] where the expression belongs, and wrap phrases with <whisper>...</whisper> or <soft>...</soft> to change delivery. Combine tags with punctuation rather than stacking them.

Picking an output format

mp3 (the default) is best for web and general playback. Use wav or pcm for editing and post-production, and mulaw or alaw at 8 kHz for telephony and IVR. Keep each request under 15,000 characters and split long scripts by paragraph.

Recommended Settings

These values were validated end-to-end during testing and make a reliable starting point:

• •voice_id: eve — clean, engaging narration for demos and announcements
• •language: en — set explicitly for deterministic output
• •codec: mp3 — broadcast-quality, ~24 kHz / 128 kbps, ideal for web
• •speed: 1.0 — natural pace; nudge up slightly only for fast reads
• •text: one to three well-punctuated sentences per request for the most natural pacing

A short, well-punctuated eve prompt produced clean, artifact-free English narration in testing, making it a dependable default before you branch into other voices or telephony formats.

How many voices does Grok Text-to-Speech offer?

Five built-in voices: ara, eve, leo, rex, and sal, with eve as the default.

What languages are supported?

Twenty languages via BCP-47 codes, plus an auto option for automatic detection.

Can I control emotion and pacing?

Yes. Inline tags such as [pause] and [laugh] and wrapping tags such as <whisper> shape delivery.

What audio formats can it output?

mp3, wav, pcm, mulaw, and alaw, with configurable sample and bit rates.

How long can the input text be?

Up to 15,000 characters per request; split longer scripts into segments.

Which voice should I start with?

Eve is a strong, engaging default for most demos and announcements.