Overview

• Nova Sonic consolidates speech-to-text, text understanding, and text-to-speech into a single system, enabling natural and context-aware communication.
• The model supports real-time, interactive voice interactions with low latency, handling interruptions and maintaining conversational flow.
• Benchmark tests show Nova Sonic outperforms competitors like OpenAI's GPT-4o and Google’s Gemini Flash 2.0 in conversational quality and accuracy.
• It achieves a word error rate of 4.2% on multilingual benchmarks and demonstrates strong performance in noisy, multi-speaker environments.
• Available now via Amazon Bedrock, Nova Sonic is nearly 80% cheaper than GPT-4o and is already integrated into Alexa+, Amazon’s upgraded voice assistant.