2026-11-04 · Fonte：Google Blog
Google's latest Gemini audio pair, Gemini 3.8 Live and Gemini 3.5 Transcribe, are available today through ai.studio/live, targeting developers building real-time voice applications.
Google has put two new voice-oriented members of its Gemini model family into developers' hands: Gemini 3.8 Live and Gemini 3.5 Transcribe. Announced on the company's blog under its developer tools coverage, the models are aimed squarely at building real-time voice applications, and Google's call to action is immediate — both can be tried today through ai.studio/live.
The pairing reads as a deliberate two-part stack. The Live designation points to real-time audio interaction — the conversational layer required for applications that respond to speech as it happens — while Transcribe names the transcription layer, turning spoken audio into text. Presented together in a single announcement about real-time voice applications, the two models give developers both the listening and the writing halves of a voice pipeline from one vendor.
Distribution carries its own message. Rather than routing developers through a separate product page, Google points them to a single address — ai.studio/live — making the studio environment the first stop for hands-on access. For teams weighing voice interfaces for production use, the on-ramp is deliberately friction-free: the company's stated invitation is to try the models today.
For the AI-compute audience, the release matters less as a headline model launch than as a marker of where the Gemini family is heading: deeper into specialized, latency-sensitive modalities. Real-time voice is among the most demanding inference workloads — continuous audio in, immediate response out — and every expansion of such models into general developer access feeds the demand picture that data-center planners and accelerator buyers track. Google has now made that workload a standard developer offering.
The published announcement is compact, naming the two models, their target use case — real-time voice applications — and the access route through ai.studio/live. Broader specifics around benchmarks, pricing and enterprise terms were not part of the disclosed material.
Tags: Google · Gemini 3.8 Live · Gemini 3.5 Transcribe · Real-Time Voice · Speech-to-Text · Google AI Studio
萬安算交所 AIXX · AIXX · Exchange verificado · Ir ao mercado