Voice · xai/grok-text-to-speech
Grok TTS — long texts voiced in a single run
Grok TTS is xAI’s text-to-speech model that takes up to 14,500 characters in a single run. Voice-over costs 2.2 ₽ per 2,000 characters. Five voices, 16 languages with auto-detection and tags for pauses, laughter and whispers right in the text make it handy for articles, chapters, instructions and courses.
from 2.2 ₽ per voice-over up to 2,000 chars
Why Grok TTS is the pick for long material
- Up to 14,500 characters per run — you have to cut a long text into pieces less often.
- Price: 2.2 ₽ per 2,000 characters, then in proportion to length.
- All five of the model’s voices: eve — upbeat, ara — warm, rex — confident, sal — soft, leo — firm.
- Tags in the text: [pause] adds a pause, [laugh] a laugh, and a fragment inside <whisper>…</whisper> is whispered.
- A short text is usually ready in a few seconds; the recording is saved to your history, so you can close the page.
The five Grok TTS voices
All five of the model’s voices are available in limko. The right column is a hint on where to start.
| Voice | Character | Good for |
|---|---|---|
| eve | upbeat | ads, announcements, social media videos |
| ara | warm | stories, podcasts, audio articles |
| rex | confident | presentations, reviews, news |
| sal | soft | fairy tales and calm texts |
| leo | firm | instructions, lessons, business texts |
Long text: length, price, workflow
An article, chapter or lesson of up to 14,500 characters can be sent in one piece, and the voice and delivery will not change midway through the text.
The price grows with length: 2.2 ₽ for every 2,000 characters, so a 10,000-character text costs five times as much. The amount to be charged is shown on the button in advance, and if something fails, the credits come back on their own.
Material longer than the limit is split by chapter or section. The parts are easy to keep in one folder, and limko’s merge tool joins them into a single track — like trimming, it is free. Before you run, “Prepare for voice-over” comes in handy: for a small separate fee the service adds stress marks and the letter ё.
Price and strengths of the text-to-speech models
Prices are per 2,000 characters, from the limko catalog. The minimum charge in any model equals 300 characters.
| Model | Price per 2,000 characters | Strength |
|---|---|---|
| Grok TTS | from 2.2 ₽ | up to 14,500 characters per run |
| Qwen3-TTS | from 3.1 ₽ | voice cloning |
| Inworld TTS | from 7.1 ₽ | directions in brackets, character lines |
| MiniMax Speech Turbo | from 18 ₽ | fast, expressive voice-over |
| MiniMax Speech HD | from 18 ₽ | studio sound quality |
| Gemini TTS | from 24 ₽ | expressive delivery, emotion tags |
Pauses, laughter and whispers right in the text
Grok’s expressiveness is set with tags inside the text: [pause] inserts a pause, [laugh] a short laugh, and the model whispers the words between <whisper> and </whisper>. For example: “The main rule [pause] <whisper>not a word to anyone</whisper>”.
For the whole text at once, limko’s ready-made delivery styles are handier — from neutral and calm to narrator voice. Choose the language from the 16 on the list or leave auto-detection on.
Frequently asked about Grok TTS
How much does voice-over in Grok TTS cost?
2.2 ₽ per 2,000 characters. A longer text costs proportionally more, and a very short one is billed as 300 characters.
What is the longest text I can voice at once?
Up to 14,500 characters per run. If the material is longer, split it into parts: merging finished recordings in limko is free.
Does Grok TTS support Russian?
Yes. The limko list has 16 languages, Russian among them; you can leave auto-detection on, and the language will be detected from the text.
How is Grok TTS different from Gemini TTS?
Grok is cheaper and takes several times more text per run. Gemini TTS costs more but is more expressive: the model receives its delivery as an instruction, and it has 12 voices in limko versus five.
Can I clone my own voice?
Not in Grok TTS — it has five preset voices. In limko, voice-over in your own voice is available only in Qwen3-TTS.