Google Gives Gemini a Face, and India's Call Centres Notice

The News

Google has begun rolling out a feature called Live Avatar for its Gemini 3.8 Live model, giving the assistant an animated on-screen persona that lip-syncs and shifts facial expressions as it talks. Reported by The Verge, the tool lets users hold a real-time spoken conversation while watching a digital face respond, rather than staring at a text box or a waveform.

The headline capability is language. Google says Live Avatar can move across the 97 languages it supports without, in the company's words, degrading video fidelity or introducing visual drift. A demonstration clip showed the avatar switching between English and Japanese with mouth movements matched to each. The persona can also surface information on screen mid-conversation.

For now the feature is restricted to Gemini Enterprise customers. It arrives roughly a week after Google unveiled the underlying Gemini 3.8 Live model, which the firm says processes visual inputs in near real-time. Organisations can pick from a library of preset avatars or build their own, and every output carries Google's invisible SynthID watermark plus what the company describes as safeguards to respect identity.

Why It Matters

Giving an assistant a face is a deliberate bet that presence, not just accuracy, is what makes people trust and use AI at work. The last time a major lab tried to make software feel convincingly human was Google Duplex in May 2018, the phone-booking demo that mimicked a person well enough to unsettle the audience and force Google to promise the bot would identify itself. Eight years on, the pitch has shifted from disguising the machine to dressing it up as a branded colleague.

That shift matters commercially. Enterprise-only availability signals where Google sees revenue: customer service, onboarding, training and sales, where a talking face can plausibly replace or augment a human agent. It also raises the stakes on synthetic-media governance, which is why the SynthID watermark and identity safeguards are being led with, not buried in a footnote.

Indian Angle

No market tests a 97-language promise harder than India. A single avatar that can move between Hindi, Tamil, Telugu, Bengali and Marathi without visual drift is aimed squarely at the country's outsourcing and enterprise-support economy, where language coverage is the whole business. If it works as shown, it pressures the very BPO revenues that Bengaluru, Hyderabad and Gurugram have long depended on, while handing domestic firms a cheaper way to serve tier-two and tier-three customers in their own tongue.

It also lands on home-grown competition. Sarvam AI and Krutrim have both staked their pitch on Indic-language voice and speech, betting that local models understand local users better. A watermarked, expression-matched avatar from Google is a direct challenge to that thesis, and to the enterprise contracts both are chasing.

Regulators will watch the face, not just the words. India's IT Rules amendments on synthetic and deepfake content, and the labelling obligations taking shape under MeitY, make an on-screen human likeness a compliance question as much as a product one. SynthID may help, but enterprises deploying a digital persona will still answer for consent, disclosure and data handling under the DPDP framework.

FAQ

Who can use Live Avatar right now?

Access is limited to Gemini Enterprise customers at launch. Google has not published a timeline for wider or consumer availability, so individual users and smaller businesses cannot switch it on yet. The feature builds on the Gemini 3.8 Live model released about a week earlier.

How many languages does it support?

Google says Live Avatar works across the 97 languages the model supports, and can transition between them without degrading video fidelity or introducing visual drift. A demo showed English and Japanese with matched lip movement, though per-language quality across all 97 has not been independently verified.

How does this affect Indian outsourcing firms?

Multilingual avatars target the language-support work that underpins Indian BPO revenue. The technology could lower costs for domestic customer service in regional languages, but it also threatens seat-based outsourcing contracts if enterprises route more interactions through a digital persona.

Is the output labelled as AI?

Yes. Google says every Live Avatar output carries its invisible SynthID watermark, alongside safeguards it describes as respecting identity. Indian deployments would additionally need to meet disclosure and consent expectations under evolving IT Rules and the DPDP framework.

Where can I read the original announcement?

The Verge's report, linked below, covers Google's rollout and the demonstration footage.

This story was reported by The Verge. Read the full original coverage at The Verge.