Details
- Google AI Developers announced Gemini 3.5 Live Translate, a new audio model focused on real-time speech translation.
- The model delivers low-latency translation across 70+ languages, aimed at live, conversational use cases.
- It processes speech as it streams, enabling developers to build applications that translate spoken language in near real time.
- Developers can experiment with the model in Google AI Studio’s live playground to test capabilities before integrating.
- Gemini Live API support and third-party integrations are available through partners including Agora, LiveKit, Pipecat, and Software Mansion.
- Official documentation and getting-started resources are provided, including technical docs and a “Start cooking” developer guide.
- Google published a dedicated blog post explaining the model, its architecture, and recommended use cases in more depth.
Impact
By bringing low-latency, 70+ language streaming translation into the Gemini stack, Google is pushing deeper into real-time communication and developer voice AI workflows. Native integrations with platforms like Agora, LiveKit, and Pipecat position Gemini 3.5 Live Translate as a first-class option for voice agents and live translation apps, intensifying competition with OpenAI and other providers focused on real-time multimodal APIs.
