Google has launched Gemini 3.8 Live with Live Avatar for enterprise customers in the U.S. and EU, enabling real-time, lip-synced video responses across 97 languages. The feature integrates speech-to-speech processing with synthetic video generation for applications such as virtual receptionists and interactive customer support.
The launch of Google Gemini 3.8 Live Avatars shifts AI interaction from text and audio into high-fidelity, low-latency video streaming. By removing the transcription step through native speech-to-speech processing, Google reduces the latency that previously hindered lifelike virtual assistants. This development places Google in direct competition with Meta in the race for real-time digital twins, specifically targeting the B2B service sector where visual branding is critical. As these tools move from internal enterprise accounts to public-facing roles, the industry should monitor how the increased cost of video-based AI sessions impacts the ROI for automated customer support compared to traditional voice-only models.
Google first previewed Live Avatar at Cloud Next 2026 in April before the September 24 general availability date, signaling a multi-month enterprise readiness period during which select customers tested the feature in production environments. The company positioned the launch as a direct extension of its Gemini Enterprise platform, requiring enterprise allowlisting for custom avatar creation from reference images. This gating mechanism suggests Google is controlling brand-safety risks carefully as it pushes synthetic video into customer-facing roles.
Google's SynthID watermarking, embedded in every frame of Live Avatar output, extends the company's broader provenance strategy that it has applied across Imagen, Veo, and Gemini image generation since 2023. The watermark operates at the pixel and audio level simultaneously, making it resistant to common transformations like cropping, compression, and color grading. For enterprise buyers evaluating Live Avatar for public-facing support or receptionist roles, SynthID provides a compliance pathway for emerging AI-content disclosure requirements in both the U.S. and EU, where the AI Act's transparency obligations take effect in stages through 2026.
On the technical side, Google's developer documentation specifies that Gemini 3.8 Live operates across bidirectional WebSockets with sub-second latency, processing speech, live camera feeds, and screen broadcasts while synthesizing 24 kHz audio and synchronized 24 FPS video. The model supports non-blocking asynchronous tool execution, meaning the avatar continues speaking while backend systems process requests, and automatically cancels pending tool calls if the user interrupts. Input video is ingested at 1 FPS JPEG frames at 768x768 resolution, with configurable media resolution settings to balance token consumption against visual detail. These specifications place Live Avatar in a distinct category from pre-rendered avatar video tools, positioning it instead as a real-time conversational agent with a visual layer.
Google has launched Gemini 3.8 Live Avatars, enabling real-time, interactive video responses for enterprise customers. By utilizing native speech-to-speech processing, the system supports 97 languages with low-latency lip-syncing. This shift from text-based AI to high-fidelity video streaming allows businesses to deploy lifelike virtual assistants for customer-facing roles while maintaining brand safety.
The system supports real-time interaction across 97 different languages.
Google integrates SynthID invisible watermarking into every frame of video and audio output to verify content and meet transparency requirements.
Yes, the system supports non-blocking asynchronous tool execution, allowing users to interrupt the AI mid-sentence, which automatically cancels pending tool calls.
Enterprises can generate custom avatars by uploading a single still photo, which requires enterprise allowlisting to ensure brand safety.
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source
