Grok TTS Text to Speech: Script to Expressive Voiceover
Last verified: October 29, 2026
Grok TTS text-to-speech works best on scripts that need a performance, not a flat read: story narration, explainer voiceovers and character lines where a pause, a laugh or a whispered aside is part of what the line means. The model reads speech tags typed straight into the text. Inline tags drop a vocal expression such as a laugh or a pause at an exact point, and wrapping tags change how a whole passage is spoken, from a whisper to singing. All of this runs on a set of expressive built-in voices.
That puts the direction in the script. The Text box holds the words and the tags together, the Voice menu decides who is speaking, and the Language menu either detects the language from what you wrote or locks it in. Changing a tag or a voice just means another run, so the working loop is short: generate a passage, listen for places where the delivery doesn't match what you meant, adjust the script and generate again.
Explore more text to speech
Grok TTS at a Glance
What goes in, the voices and languages you can pick, and what each run gives you.
Input
A text prompt of up to 15,000 characters
Voice
Eve — energetic, Ara — warm, Rex — confident, Sal — smooth or Leo — authoritative
Language
21 options
Output
1 audio per run
How Do You Make Grok TTS Sound Natural?
Pauses Placed Where a Thought Should Land
One Tag Plus the Right Punctuation, Not a Stack of Tags
Delivery Styles Wrapped Around Whole Phrases
How Do You Turn a Script Into Speech With Grok TTS?
Four steps take you from a written script to a voiceover you can download.
Write your script in the "Text" box
Type or paste the words to be spoken into "Text". It is the one required field and holds up to 15,000 characters. Speech tags go in the same box, right where they should take effect. "Enhance Prompt" under the box suggests fuller versions of a rough idea and changes nothing until you pick one. Read a suggestion before accepting it, because every word you keep will be spoken.
Choose a "Voice" and a "Language"
"Voice" offers Eve — energetic (the default), Ara — warm, Rex — confident, Sal — smooth and Leo — authoritative. Match the voice to the job: Leo or Rex for a tutorial, Ara for a calm support message, Eve for an upbeat promo. "Language" defaults to Auto, which detects the language from your text. It also lists Arabic (Egypt), Arabic (Saudi Arabia), Arabic (UAE), Bengali, Chinese, English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese (Brazil), Portuguese (Portugal), Russian, Spanish (Mexico), Spanish (Spain), Turkish and Vietnamese. Pick the exact regional variant when your audience will hear the difference.
Click "Generate"
Start with a short passage that contains your trickiest tags, so you hear how the chosen voice handles them before you commit the full script. Once that passage sounds right, paste the whole script and generate again. Changing a tag or the voice and generating again is the quickest way to compare takes.
Download the result
Each run returns one audio file. Listen all the way through before you download, paying attention to the tagged moments, and keep the take you want. For a version in another voice or language, change that setting and generate again; each version is its own run.
Choosing Between Grok TTS and Gemini-TTS on Google Cloud
Where each one fits when you're deciding how to voice a script.
| Criterion | Grok TTS | Alternatives | Best For |
|---|---|---|---|
| Language coverage | xAI documents twenty languages for Grok TTS, with separate regional variants for Arabic, Portuguese and Spanish. | Google documents Gemini-TTS on Google Cloud as supporting more than 70 locales. | If your audience's language is on Grok's list, you can stay with Grok TTS; for a language outside it, Gemini-TTS's wider locale coverage is the one to check. |
| Where you run it | Here it runs from a short form where you enter a script in the Text box, choose a Voice and a Language, then generate and download the audio. | Gemini-TTS runs inside Google Cloud, either through the Vertex AI API or in Media Studio. | Creators who want to go from script to audio file in one place can use the form here, while teams whose pipeline already lives on Google Cloud can keep speech generation inside it. |
| Long scripts | xAI describes Grok TTS's standard request as the way to turn long-form text into speech. | Google documents Gemini-TTS for speech ranging from short snippets to long-form narratives. | Script length doesn't separate the two, so let language and workflow decide; with Grok TTS here, a script longer than the Text box allows goes in as several runs. |
Fixing Common Grok TTS Problems
What to change when a read doesn't come out the way you planned.
My script is too long to fit in the Text box.
The voice reads my text in the wrong language or accent.
I want to hear my script in several voices, but I only get one file.
Frequently Asked Questions
How many languages does Grok TTS support?
xAI documents 20 languages for Grok TTS, and each one can be chosen in the Language menu alongside automatic detection. The only Chinese variant xAI lists is Simplified Chinese.
Can Grok TTS laugh, whisper or sing?
Yes, through speech tags typed into the Text box. Inline tags produce a sound at one point in the text: [pause], [long-pause], [hum-tune], [laugh], [chuckle], [giggle], [cry], [tsk], [tongue-click], [lip-smack], [breath], [inhale], [exhale] and [sigh]. Wrapping tags are written in angle brackets around a passage and change how it is delivered. Their names are soft, whisper, loud, build-intensity, decrease-intensity, higher-pitch, lower-pitch, slow, fast, sing-song, singing and emphasis.
What makes one Grok TTS generation cost more than another?
The credit cost of a run is worked out from the options you choose and appears above the Generate button once you're signed in. What changes it is the length of the text you enter in the Text box, so a short test passage and a full script will show different costs.
Do I need an API key or code to use Grok TTS?
No. xAI offers Grok TTS to developers through a REST API and a WebSocket API for real-time speech. Here, the model runs behind a form: you enter your script, choose a voice and a language, generate, and download the audio.