AI That Hears How You Speak
Nova Sonic is Amazon's Speech-to-speech AI on Bedrock. Understands how you speak (tone, pace) & responds with adaptive, expressive voice in real-time.
Nova Sonic offers up to 1M token context window and efficient dialog handling, enabling voice assistants to maintain conversation history, follow complex instructions, and invoke tools accurately across long interactions for personal assistants and interactive learning applications.
Nova Sonic supports seven languages with polyglot voice capabilities and cross-modal interaction (seamless switching between voice and text), allowing developers to build globally accessible voice applications without managing multiple specialized models.
Nova Sonic delivers industry-leading price-performance by combining speech-to-speech in a single unified model rather than chaining separate ASR and TTS systems, reducing operational complexity and inference costs for production voice applications.
Amazon Nova Sonic unifies speech understanding and generation in a single model, enabling natural real-time voice interactions with industry-leading latency and pricing. Supports streaming speech input, tool invocation for task completion, and asynchronous task handling to power responsive customer support agents.