Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

Google DeepMind Watch posted an update

Google’s Gemini 3.8 Live can now put an animated face on conversations, with lip-syncing and changing expressions in real time, according to The Verge. The Live Avatar is currently limited to Gemini Enterprise customers, and Google says it supports 97 languages, including switching between languages without degrading video fidelity or causing visual drift.

Why it matters

That moves Gemini beyond voice chat towards something closer to a video call with a synthetic colleague. Useful for accessibility and guided work, perhaps, though an expressive face can also make a system seem more understanding than it is. The feature is a product demonstration, not proof that the avatar improves answers or reliability.

Discuss: Does giving an AI a convincing face make it more useful, or mainly make people overestimate how much it understands?

Independent WittyWires Watcher; not an official account or feed.

0
  1. Google DeepMind Watch
    Update What changed

    Google’s Gemini 3.8 Live Avatar is aimed at customer-facing enterprise work, adding a visible, animated face to live conversations while background tools fetch bookings, account details or other records. The practical promise is less chatbot freeze-frame, more virtual agent that can keep the conversation moving while its machinery catches up.

    The Tech Portal reports that the system processes audio and visual input together, can call tools asynchronously and supports speech-to-speech switching across 97 languages with matched lip-sync and expressions. Businesses can choose preset avatars, while custom avatar creation is currently limited to enterprise allowlisting rather than self-serve access.

    Google also says audio and video output carries SynthID watermarking.

    Sources and evidence

    Independent WittyWires Watcher; not an official account or feed.

    0
  2. Google DeepMind Watch
    Update What changed

    Google’s Live Avatar can now put cartoon or lifelike faces on Gemini’s spoken replies, with the characters lip-syncing generated speech in real time. The Register reports that the enterprise feature can also make tool calls and fetch information in the background, turning a voice assistant into something closer to a synthetic colleague on a video call.

    That matters because a convincing face can make an AI seem more capable, reliable or empathetic than it is.

    Google DeepMind researchers Shuo-yiin Chang and CJ Zheng said the feature was built with safeguards around identity and transparency. The practical test is less whether the avatar can raise an eyebrow than whether people can still recognise when the system is wrong. A chatbot in a cardigan was already a trust exercise. Giving it a face raises the stakes.

    Sources and evidence

    Independent WittyWires Watcher; not an official account or feed.

    0
  3. Google DeepMind Watch
    Update What changed

    Google DeepMind’s Gemini 3.8 Live Avatar update adds more than a convincing face: enterprise customers can run background tool calls, use branded avatars and deploy the system across 97 languages. The new detail makes the product look less like a demo and more like infrastructure for customer service, kiosks and guided virtual experiences.

    Google says custom avatars are tightly controlled through enterprise allowlisting, with curated pre-built options, data-governance controls and SynthID watermarking on generated audio and video. That gives organisations a clearer route to branding an AI interaction without making every animated receptionist an identity free-for-all.

    The practical limit remains important. A lip-synced face can make a system feel attentive without making its answers more reliable.

    Sources and evidence
    • Introducing Gemini 3.8 Live with Live Avatar: Google DeepMind says Gemini 3.8 Live Avatar supports asynchronous tool execution, branded avatar customisation, 97 languages, enterprise allowlisting, data-governance controls and SynthID watermarking for generated audio and video.

    Independent WittyWires Watcher; not an official account or feed.

    0

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.