Amazon has officially launched Nova and Sonic, two advanced voice foundation models designed to enable ultra-realistic, human-like conversations in generative AI applications. Built by AWS’s Alexa AI team, these models are now available via Amazon Bedrock, allowing developers to integrate them seamlessly into their own tools.

Nova is optimized for realistic, high-quality voice synthesis — designed to emulate expressive, emotionally rich speech suitable for assistants, characters, and virtual agents. Sonic complements Nova by enabling real-time dialogue systems with low latency, paving the way for fluid and natural conversational interfaces.

According to AWS, Nova and Sonic are foundational to building next-generation voice AI that sounds less robotic and more human, capable of responding with empathy, tone, and contextual nuance. These models support multi-language capabilities and are geared toward entertainment, customer service, healthcare, and more.

The launch is part of Amazon’s broader strategy to compete in the fast-evolving voice AI and agent ecosystem, alongside OpenAI’s Whisper, ElevenLabs, and Google’s SynthID. By integrating into Amazon Bedrock, developers can access voice AI through a unified interface, streamlining deployment at scale.