introducing gemini 3.5 live translate, our latest audio model: - low-latency translation across 70+ languages - auto-detection for multilingual inputs in a single session - native audio processing th
The Hook
“Hey everyone. Today we're excited to launch Gemini 3.5 Live Translate, our latest model built for seamless speech-to-speech translation. While it's already been powering experiences in Google Translate, we are now bringing this capability directly to developers via our API. You can integrate real-time translation across 70+ supported languages for both input and output starting today. We're going to show you what it can do with two demos.”
Product Reveal
“(No spoken dialogue — ambient audio branding chime)”
Feature Teaser
“Okay, so for our first demo, we're going to show you a live dubbing experience. Here we're using the API to stream translated audio directly from a tab. Watch as we listen to the Google I/O keynote in Hindi.”
Feature Teaser
“Now, the second demo is for live translating talks and presentations. Let's say you're presenting to an international audience that speaks a range of different languages. Here as the speaker, I can create a new session and I can capture the audio from my microphone. I then get a URL or a QR code for my attendees to scan this on their phone and join the session and start listening in their own preferred language. So, let's start maybe with Japanese.”
Feature Teaser
“Now I'm switching to German. The model recognizes the language switch automatically. I don't have to configure anything. I think when I speak in Sinhala, what I say in Sinhala is translated into English without any configuration. And listen to how natural the translation sounds. No choppiness, no artificial pauses. It flows like a completely natural language.”
Call to Action
“The model is now available on the Gemini API and in AI Studio. You can also try it in Google Translate on iOS and Android by connecting any pair of headphones. And we're also rolling it out in private preview on Google Meet. We can't wait to hear what you build.”
Synthesizable in Remotion · key components of the developer-live-demo-broadcast template.
Expanding pill shape with multi-color radial rim glow cycling multilingual title text.
Compose an absolute positioned div with border-radius 9999px, animated CSS linear-gradient border stroke, and spring-driven opacity/scale transforms.
Real-time streaming text block appending translated tokens with smooth auto-scrolling.
Use Remotion Sequence with slice-based character reveals and spring-animated translateY offset for newly incoming subtitle lines.
Rounded portrait webcam feed hovering in screen corners with subtle drop shadows.
Wrap video feeds in an AbsoluteFill container with fixed width, border-radius 16px, backdrop-filter blur, and smooth spring-based enter/exit position transitions.
Purpose. Introduce the new model and its developer availability.
Execution. Direct-to-camera dual presenter framing in a modern studio.
Purpose. Establish product identity and multi-language breadth.
Execution. Animated glowing gradient capsule cycling multiple scripts.
Purpose. Prove low-latency speech-to-speech dubbing from live media.
Execution. Google AI Studio playground screen capture streaming Hindi audio/text.
Purpose. Demonstrate scalability across multiple simultaneous output languages.
Execution. Multi-card listener interface expanding from a single broadcast microphone feed.
Purpose. Showcase automatic input language switching without re-configuration.
Execution. Switching between German, Sinhala, and English live on camera.
Purpose. Deliver clear call-to-action for developers and end-users.
Execution. Return to presenters followed by animated logo lockup.