Thursday, September 24, 2026
Tech Beat
Sep 24, 2026, 4:20 PMArtificial Intelligence

Gemini 3.8 Live Avatar Brings 97-Language AI Video to Enterprise

Gemini 3.8 Live adds low-latency video avatars, 97-language speech, background tool calls and SynthID safeguards for Gemini Enterprise users from Sept. 24.

Listen to this briefingAudio briefing

Summary

Gemini 3.8 Live with Live Avatar launched in Gemini Enterprise on September 24, 2026, one week after Gemini 3.8 Live. It combines native live dialogue with low-latency video generation, processing visual and audio inputs together to listen, see and respond through animated personas with synchronized speech, expressions and turn-taking. Intended uses include customer service and interactive walkthroughs, with preset characters offering distinct appearances, voices and expressive styles.

Asynchronous tool calling lets an avatar fetch data or perform tasks, such as hotel guest check-in, without pausing conversation. Native multilingual speech-to-speech synchronization supports mid-conversation switching across 97 languages, dynamically adjusting lip movements and expressions without reduced video fidelity or visual drift. Enterprises can generate responsive custom avatars from a reference image while retaining likeness, styling or character identity, although creation requires allowlisting. Every generated audio and video output carries an imperceptible SynthID watermark to support detection and limit misinformation or misattribution. API documentation and a safety model card are available, but pricing, measured latency and access beyond Gemini Enterprise were not disclosed.

Positives

  • 97-language speech-to-speech synchronization adapts lip movements and expressions during mid-conversation language changes without reducing video fidelity.
  • Asynchronous tool calling lets Live Avatar fetch data and complete tasks while maintaining uninterrupted dialogue.
  • Visual and audio inputs are processed simultaneously, enabling animated responses with synchronized speech, expressions and turn-taking.
  • Reference images can become responsive custom avatars while preserving likeness, brand styling or character identity.
  • SynthID invisibly watermarks every generated audio and video output to improve detection of AI-created content.

Risks & concerns

  • Custom avatar creation is restricted to enterprises admitted through allowlisting.
  • Gemini 3.8 Live with Live Avatar is available only through Gemini Enterprise.
  • Precise latency measurements, pricing and plans for access beyond Gemini Enterprise remain undisclosed.
  • Realistic generated personas carry misinformation and misattribution risks, which SynthID is intended to reduce rather than eliminate.
Primary sourceGoogle DeepMind Newshttps://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-with-live-avatar/
Read full article
Editorial note: Tech Beat summarizes and analyzes third-party reporting. The source link is the authoritative article. This page does not reproduce the full source text.

More From The Wire

Artificial IntelligenceSep 24

Google Project Suncatcher Satellite to Test Orbital AI on October 1

Artificial IntelligenceSep 24

Shield AI, Waabi and GM to Detail Safety Tests for Physical AI at Disrupt 2026

Artificial IntelligenceSep 24

Lovable ARR Tops $600M as Enterprise Vibe Coding Surges