28 named Grok voices
From Eve, energetic with a British accent, to Orion, rich and cinematic, and Carina, soft and soothing. Play each voice's official sample before you spend a credit.
Updated for the October 14, 2026 release
Grok Voice 3.0 is xAI's new speech-to-speech model for real-time voice agents. This page explains what changed and how to get access, and gives you a text to speech tool that reads your own script in the Grok voices, such as Eve, Ara and Leo.
20 free credits when you sign in. No card needed.
Paste a script, pick a voice and a language, add tags like [laugh] or <whisper>, and download the MP3. This tool runs xAI's Grok text to speech model: it reads your text aloud, it does not hold a live conversation.
Speech tags: click to insert at the cursor
Sign in with Google and get 20 free credits.
Your audio will appear here
1 credit covers 100 characters, rounded up. The cost updates as you type.
Grok Voice 3.0 is the third generation of xAI's speech-to-speech model. xAI announced it on October 14, 2026 and made it available in the xAI Voice Agent API. Speech-to-speech means the model hears a person talking and answers out loud, in real time, so it is built for voice assistants, phone agents and other live conversations.
xAI calls it the world's most intelligent speech-to-speech model. Its launch post reports a score of 86.9 on the Artificial Analysis Speech-to-Speech Index, first place and up from 81.3 for Grok Voice 2.0, with latency under one second. The price stayed at $0.08 per minute ($4.80 per hour), the same as Grok Voice 2.0, and voice agents using the grok-voice-latest model alias were moved to 3.0 automatically.
According to xAI, Grok Voice already handles millions of conversations a day across Tesla and SpaceX. It powers Grok Bot voice mode and Connectors in Tesla vehicles, and Starlink call centers use it to answer about 20,000 customer calls a day in several languages, with about 70% resolved without a human.
| Benchmark | Grok Voice 3.0 | Grok Voice 2.0 |
|---|---|---|
| Artificial Analysis Speech-to-Speech Index | 86.9 | 81.3 |
| Big Bench Audio (speech reasoning) | 98.5% | 97.2% |
| Speech Agent Arena Elo (human preference) | 1070 | 1011 |
| Task success rate | 94.8% | 94.6% |
| Full Duplex Bench (turn-taking) | 97.0% | 95.1% |
| τ-Voice (support calls with noise and interruptions) | 71.6% | 56.5% |
| Voice Agent API price | $0.08 / min | $0.08 / min |
Figures as published in xAI's launch post, which cites Artificial Analysis results dated October 13, 2026. Source: x.ai/news/grok-voice-3
There are three official ways to reach Grok Voice 3.0 today, plus one shortcut if what you really need is a voiceover.
xAI offers a free trial through its developer console at console.x.ai. The playground there lets you talk to Grok Voice before you write any code.
Developers connect over WebSocket to wss://api.x.ai/v1/realtime with the grok-voice-latest model, which now points to Grok Voice 3.0. Usage is billed at $0.08 per minute of conversation.
xAI says Grok Voice 3.0 powers Grok Bot voice mode in Tesla vehicles, where drivers ask for things like ordering a coffee or booking dinner without touching the screen.
Grok Voice 3.0 is made for live, two-way calls. If you want a finished audio file read in a Grok voice, paste your script into the generator above and download the MP3. xAI's text to speech API uses the same voice names as its voice agents; Eve is the default in both.
Every voice below is available in the generator, and each one can speak every supported language. Press play to hear xAI's own sample of the voice. The tone notes come from xAI's voice list.
Tags change how a line is delivered: a laugh, a sigh, a pause, a whisper or slower speech. These clips from xAI's documentation show a few of them. Type the same tags into your script to use them.
Voice and tag samples: xAI documentation (docs.x.ai). Grok and xAI are trademarks of their owner.
One text box, the full Grok voice line-up and clear per-character pricing.
From Eve, energetic with a British accent, to Orion, rich and cinematic, and Carina, soft and soothing. Play each voice's official sample before you spend a credit.
English, Chinese, Spanish, French, German, Japanese, Korean, Hindi, Arabic and more. Every voice speaks every supported language, and auto detect reads the language from your text.
Write [laugh], [sigh] or [pause] where a sound should happen, or wrap words in <whisper>, <slow> or <emphasis> to change how they are said.
Enough for a full video script, a podcast intro or a short lesson. Split longer projects into parts and keep the same voice for each part.
Each file is a 24 kHz, 128 kbps MP3 saved to your history. Use it in videos, ads, courses and podcasts, commercial projects included.
1 credit covers 100 characters, and the exact cost shows before you click. If a generation fails, the credits go back to your balance automatically.
A few starting points, with voices from xAI's list that suit each job.
Write the script, generate it once, and drop the MP3 under your edit. Upbeat voices like Eve, Helios or Iris keep short videos moving.
Orion is listed as rich and cinematic and Zagan as powerful and dramatic, a good fit for trailers, fiction and documentary-style narration. Add [pause] and <slow> for pacing.
Altair and Perseus are listed for advertising. Keep the copy short, put the product name in <emphasis>, and test two voices before you pick one.
Luna, Cosmo and Lumen are listed for education. Turn a lesson plan or help article into audio learners can play on the go, in any of the 20 languages.
Record IVR prompts, hold messages or voicemail greetings with calm voices like Carina, Celeste or Aurora, and regenerate them whenever the wording changes.
Helix and Kepler are listed for podcasts and commentary. Draft an intro, an ad read or a recap in minutes before you record the real thing.
1 credit = 100 characters of text. The exact cost of each script shows before you generate, and failed runs are refunded automatically.
Basic
For occasional voiceovers
Pro
For weekly videos and podcasts
Studio
For busy creators and small teams
FAQ
Short answers about Grok Voice 3.0, this site and how credits work.
Grok Voice 3.0 is xAI's speech-to-speech model for real-time voice agents. It was announced on October 14, 2026, costs $0.08 per minute in the xAI Voice Agent API, and replaced Grok Voice 2.0 for everyone using the grok-voice-latest alias.
No. Grok Voice 3.0 is a live conversation model that listens and answers in real time. The generator here uses xAI's Grok text to speech model, served through fal.ai, to read a script you write in the Grok voices. Use it when you need a finished audio file rather than a voice agent.
No. grok-voice.net is an independent site and is not affiliated with or endorsed by xAI. For xAI's own console, API keys and documentation, go to x.ai and docs.x.ai.
Yes. New accounts get 20 free credits after signing in with Google, enough for about 2,000 characters of speech, and the daily check-in adds a few more. You can play every voice sample on this page without an account.
1 credit covers 100 characters, rounded up, so a 1,500-character script costs 15 credits. Monthly plans start at $9.90 for 1,500 credits, and one-time packs start at $9.90 for 1,000 credits that never expire.
Type an inline tag such as [laugh], [chuckle], [sigh], [breath] or [pause] where you want the sound, or wrap words in tags like <whisper>…</whisper>, <slow>…</slow> or <emphasis>…</emphasis>. The buttons under the text box insert them for you. Tags count toward your characters.
English, Chinese, Spanish (Spain and Mexico), French, German, Italian, Portuguese (Brazil and Portugal), Japanese, Korean, Hindi, Bengali, Indonesian, Vietnamese, Turkish, Russian and Arabic (Saudi Arabia, UAE and Egypt), plus auto detect. Choosing a language does not translate your text.
Yes. You can use the audio generated for you in videos, ads, podcasts, courses and other commercial work, as long as you have the rights to your script and you don't use it to impersonate a real person. The Terms of Service have the details.
The credits for that run go back to your balance automatically. Your scripts and audio stay private in your history, and we don't use them to train AI models.
Sign in with Google, get 20 free credits and turn your first script into an MP3 with Eve, Leo, Orion or any of the 28 voices.