Amazon Nova foundation model
Amazon Nova Sonic
Amazon Nova speech-to-speech model for real-time conversational voice applications through Amazon Bedrock streaming APIs.
Key points
- Unifies speech understanding and speech generation in a single model.
- Supports real-time conversational voice interactions.
- Uses an Amazon Bedrock bidirectional streaming API.
- Is the Nova model for speech; Nova understanding models do not accept audio or speech input.
When to use it
- Choose Sonic for natural-sounding voice assistants and conversational voice apps.
- Use it when the application needs real-time speech-to-speech interaction.
- Use through Amazon Bedrock bidirectional streaming for voice conversations.
Exam tips
- Sonic is the Nova model for speech-to-speech.
- Do not choose Micro, Lite, Pro, or Premier for audio or speech input.
- Canvas, Reel, and Sonic map to image, video, and speech generation respectively.