Amazon Unveils Nova Sonic: A Next-Generation Gen AI Voice Model for Natural, Real-Time Conversations

Amazon introduced *Nova Sonic*, a new foundation model that unifies speech understanding and generation to create more natural, accurate, and engaging voice-powered applications. Available through Amazon Bedrock, Nova Sonic enables real-time AI agents and voice apps for industries such as healthcare, education, and travel, eliminating the need for multiple speech and language models.

With significantly improved latency, lower cost, and support for expressive, native voices in various accents, Nova Sonic is already being used by companies like ASAPP, Education First, and Stats Perform to enhance user interactions. According to benchmark tests, Nova Sonic outperforms rivals such as OpenAI’s GPT-4o (Realtime) and Google Gemini Flash 2.0 in accuracy, quality, and speed, delivering responses in just 1.09 seconds and reducing word error rates across multiple languages.

Developers can leverage its fast inference, built-in transcript generation, and seamless tool integration for more intelligent, conversational AI agents—paving the way for voice applications that are as functional as they are human.