Gemini Live
谷歌推出的实时AI对话产品，支持摄像头实时视觉理解、图像生成和第三方服务联动，面向全球用户免费开放
Core Facts
Timeline (last 90 days)
In episode 47 of the Chain of Thought podcast, Paige Bailey, head of AI developer relations at Google DeepMind, demonstrated Stanford's open-source Pupper robot using a Raspberry Pi and the Gemini API for vision, with Gemini Live supporting text-to-speech and robot actions such as handshaking and following.
In the demo, Gemini Live triggers the Lyria 3 music generation model via function calling, composes live in the style specified by the user, and generated complete lyrics and melody.
ExamReady's speaking practice feature implements a real-time conversation partner based on Gemini Live, with each speaking task costing approximately ten cents
In a test drawing an SVG of a pelican riding a bicycle, the standard native audio model produced crude results, but after switching to the max advanced reasoning model, the generated SVG included a helmet, scarf, and a richly detailed bicycle structure.
The background reasoning mechanism decouples time-consuming reasoning computation from the real-time audio stream, allowing the model's voice response layer and reasoning computation layer to run in parallel
This update is the first to open up advanced reasoning (high reasoning) and background reasoning capabilities in Gemini Live
Google AI Studio launched a Gemini Live version supporting switching among up to 97 languages and real-time visual understanding, with its Extended Thinking capability enabling reasoning while speaking and invoking tools to complete tasks.
Initializing the Live model requires passing in the model name, voice role configuration, response modality, and the set of tools to be used.
All Facts (10)
Gemini Live supports three multimodal interaction features: real-time voice conversation, real-time camera recognition, and screen-sharing collaboration.
65%UnverifiedGemini Live的实时交互依赖流式推理技术，配合谷歌自研的TPU芯片集群实现毫秒级延迟
75%UnverifiedThe background reasoning mechanism decouples time-consuming reasoning computation from the real-time audio stream, allowing the model's voice response layer and reasoning computation layer to run in parallel
60%UnverifiedExamReady's speaking practice feature implements a real-time conversation partner based on Gemini Live, with each speaking task costing approximately ten cents
50%UnverifiedIn a test drawing an SVG of a pelican riding a bicycle, the standard native audio model produced crude results, but after switching to the max advanced reasoning model, the generated SVG included a helmet, scarf, and a richly detailed bicycle structure.
50%UnverifiedInitializing the Live model requires passing in the model name, voice role configuration, response modality, and the set of tools to be used.
50%UnverifiedIn episode 47 of the Chain of Thought podcast, Paige Bailey, head of AI developer relations at Google DeepMind, demonstrated Stanford's open-source Pupper robot using a Raspberry Pi and the Gemini API for vision, with Gemini Live supporting text-to-speech and robot actions such as handshaking and following.
50%UnverifiedIn the demo, Gemini Live triggers the Lyria 3 music generation model via function calling, composes live in the style specified by the user, and generated complete lyrics and melody.
50%UnverifiedThis update is the first to open up advanced reasoning (high reasoning) and background reasoning capabilities in Gemini Live
60%UnverifiedGoogle AI Studio launched a Gemini Live version supporting switching among up to 97 languages and real-time visual understanding, with its Extended Thinking capability enabling reasoning while speaking and invoking tools to complete tasks.
50%