Voice Agent and TTS

Updated November 3, 2026. Based on the Pricing, Rate Limits, Text to Speech pages and xAI release notes: voice agent at $0.08 per minute (not $0.05), TTS at $15 per million characters (not $4.20), simultaneous sessions by tier (not 100, nor 30 minutes per session), TTS available since March 2026; tiers apply automatically to voice models, Imagine requires contacting sales. Sources: Pricing · Rate limits · Release notes.

Grok’s Audio Capabilities

xAI offers two audio services via its API: the Voice Agent (Speech to Speech API), available since December 2025, for real-time voice conversations, and Text-to-Speech (TTS), available since March 2026, for voice synthesis. These services let you build complete voice conversational interfaces, from automated phone systems to embedded voice assistants.

Voice Agent API

The Voice Agent API lets you create bidirectional real-time voice conversations. The user speaks, the model listens, understands, and responds orally—all with latency low enough for natural conversation.

Key Features

• Model: grok-voice-3.0, which the grok-voice-latest alias points to
• Simultaneous sessions: 10 at the entry tier, then 20, 50, 100, and 200 by tier
• Pricing: $0.08 per minute, or $4.80 per hour, plus $0.004 per text input
• Real-time: latency optimized for natural conversation

Use Cases

The Voice Agent is designed for scenarios where voice interaction adds value over text:

• Automated customer service — phone greeting, call routing, answers to frequent questions
• Embedded voice assistants — in mobile apps or IoT devices
• Voice tutoring — educational support via voice
• Accessibility — voice interface for users who cannot or prefer not to type

Cost Considerations

At $4.80 per hour, the Voice Agent is economical for short, occasional interactions. But the cost can become significant for high-volume services. A call center handling 1,000 hours of conversation per month would represent a $4,800 budget.

The tier system also applies to voice models, just like text models: the tier depends on your cumulative spend and never goes down. To raise Imagine limits, however, you must contact xAI sales.

Text-to-Speech (TTS)

The TTS service converts text to speech with high-quality synthetic voices. It is billed by the volume of text converted.

Pricing and Limits

To estimate your costs: a 5,000-character blog post would cost about $0.075 to convert to audio. A 500,000-character audiobook would cost about $7.50.

Available Voices

xAI offers 28 built-in voices (as of November 3, 2026), common to both TTS and the Voice Agent, with various tones and uses, as well as custom voices created from a short clip. You select the voice at the time of the API request, which lets you adapt the voice to the context—a professional voice for a report, a warm voice for a tutorial.

TTS Applications

• Automated podcasts — transform your blog posts into audio content
• Narration — add an audio layer to your educational applications
• Accessibility — make text content accessible to visually impaired users
• Voice notifications — spoken alerts and reminders in applications
• Multilingual content — generate audio versions in different languages

Voice Agent vs TTS: Which Service to Choose?

The distinction is simple:

• Voice Agent: for interactive real-time conversations. The user speaks, the model responds. It is bidirectional.
• TTS: for one-way conversion of text to speech. No interaction, just voice synthesis.

If you are building a conversational assistant, you will need the Voice Agent. If you simply want to read text aloud, TTS is sufficient.

Integration into an Existing Architecture

Grok’s audio services integrate into the same API as the text and image models. You use the same API keys, the same billing system, and the same dashboard on console.x.ai.

This unification significantly simplifies architecture: a single project can use Grok 4.5 for reasoning, grok-build-0.1 for code, Grok Imagine for visuals, and the Voice Agent for oral interaction, all with a single developer account.

Pricing verified November 3, 2026 — prices change regularly: before any budget calculation, check the current rate card on the official xAI pricing page.

Key Takeaways

• The Voice Agent costs $0.08/min ($4.80/hr) for real-time conversations
• TTS costs $15 per million characters for voice synthesis
• Voice Agent: 10 to 200 simultaneous sessions by tier
• TTS: 100 concurrent requests
• Voice Agent = bidirectional, TTS = unidirectional
• Voice/Imagine tier upgrades require a manual request