#Voice AI

Tencent’s Hy-SimulMT reports roughly three seconds of translation lag. Understand the metric, Chinese-English scope, TTS dependency and deployment limits.

OpenAI now creates a custom voice from a text description for free. Where prompt-made voices work, the 100-voice limit and what it costs to use one.

Fish Audio's Drama 3 takes stage directions in brackets, edits single lines in place and ships open weights. Price, licence limits and who it suits.

Compare voice AI billing without mixing characters, tokens and minutes. Build a useful workload estimate that includes retries, silence and accepted audio.

Test a voice agent on names, addresses, prices and reference numbers. Separate wrong text from wrong speech, then verify dictionary fixes in real calls.

Tavus's Griffin-Lite joins OpenAI's GPT-Live-1 and three open models that listen while they speak. Latency, price, license and how to test each one yourself.

ElevenLabs' Eleven v4 tops Artificial Analysis' voice arena and v4 Turbo claims ~100ms. The two latency figures, the vendor tests and the v4 price.

Artificial Analysis scores realtime voice systems two ways. The arena system with the best reasoning and turn-taking scores ranks last on listener preference.

92 text-to-speech models ranked by blind listener vote, with the price per million characters, open-weight licences and the job each leader fits.

Gemini 3.8 Flash TTS is generally available with voice cloning from a 30-second sample. The promotional price ends December 31 and doubles on January 1, 2027.

GPT-Live-1 separates live conversation from backend work. Compare delegation modes, duration pricing and pilot checks before choosing a business voice setup.

Keep voice conversations useful during slow tool calls. Design progress updates, corrections and late-result handling around the task’s verified backend state.

Measure voice agent latency from speech to verified results. Use a timing dictionary to separate first audio, tool delays, playback and interruption recovery.

Three voice moves landed in one week. How to assemble a duplex layer, a transcription layer and CRM write-back when the newest model has no public API.

A vendor's 1.5-2x accuracy claim can run on its own unpublished harness while independent leaderboards rank the same models very differently. How to tell.

ChatGPT Voice arrived on Mac and Windows on July 23, 2026, wired into Codex and ChatGPT Work. The developer API is still signup-only, so this is app-only.

Claude voice mode now reaches Opus and Sonnet on paid plans, with Gmail, Calendar, Docs and Slack connectors working aloud. Still beta and turn-based.

Qwen-Audio-3.0-TTS-Plus tops the Artificial Analysis TTS arena (~1,236 Elo) at ~$27.59/1M chars, a third of ElevenLabs — but it is hosted-only, no weights.

OpenAI's GPT-Live launched July 8 with full-duplex voice that listens while it talks and delegates hard questions to GPT-5.5. The developer API is signup-only.

xAI shipped a no-code Voice Agent Builder on July 1, 2026: a plain-language prompt to a working phone agent in about two minutes, on one $0.05/min meter.

Compare voice AI agent platforms for business in 2026. ElevenLabs, Vapi, Retell, and Bland tested on latency, pricing, and enterprise CX automation.

Master Alibaba's Qwen App Beta with deep research, AI coding, voice calls, and AI camera. Complete guide to China's AI super app strategy and ecosystem.

Deploy voice AI with current ElevenLabs and Hume context, realistic ROI ranges, platform caveats, implementation steps, and measurement guidance.