Introduction

Natural conversation: Voice interactions of remarkable quality, more appropriate expressivity, and prosody (patterns of rhythm), delivered with very low latency so you can converse fluidly. Style control: Using natural language prompts, you can adapt the delivery within the conversation, steering it to adopt specific accents, produce a range of tones and expressions and even whisper. Tool integration: Gemini 2.5 can use tools and function calling during dialog. This allows it to incorporate real-time information from sources like Google Search or use custom developer-built tools, making conversations more practical. Conversation context awareness (proactive audio): Our system is trained to discern and disregard background speech, ambient conversations and other irrelevant audio, responding when appropriate. Basically, it understands when not to speak. Audio-video understanding: With native support from streaming audio and video, Gemini 2.5 can converse with you about what it sees in a video feed or through screen sharing. Multilinguality: Converse in any of our 24+ supported languages, or even easily mix languages within the same phrase. Affective dialog: Gemini 2.5 responds to the user's tone of voice, recognizing that the same words spoken differently can lead to very different conversations. Advanced thinking dialog: Gemini’s reasoning capabilities can enhance its conversation, leading to overall better performance across all features. This leads to more coherent and intelligent interactions, particularly for complex reasoning tasks.