xAI expands Grok Voice API: 21 multilingual voices, voice cloning, and expressive speech tags

What's new

• 21 new Grok Voice voices: Added to the original five, all natively multilingual (25+ languages: English, Hindi, Spanish, French, German, Japanese, Korean, and more).
• Voice cloning: Clone your own voice from a reference clip up to 120 seconds. Cloned voices inherit all TTS features including speech tags and multilingual output.
• Speech tags: Control speech delivery with [pause], [whisper], [emphasis], laughter, breathing, volume, pitch, tempo, and singing tags.
• API cost tracking: Every API response now includes a cost_in_usd_ticks field with the exact cost of the request.
• API coverage: New voices available via Voice Agent API, Text to Speech API, and Grok Voice Agent Builder.

Why it matters

With this update, Grok Voice becomes a serious competitor to OpenAI TTS and ElevenLabs. Voice cloning without excessive onboarding and expressive speech tags for nuanced delivery opens new possibilities for building production voice AI agents.

How to try it

Documentation: docs.x.ai/developers/model-capabilities/audio/voice. Voice cloning: send a reference audio clip to the /v1/voice-clone endpoint, receive a voice_id for TTS calls.