Deliver hyper-personalized viewer experiences with an agentic AI movie assistant using Amazon Bedrock AgentCore and Amazon Nova Sonic 2.0
New AI assistant understands your mood and answers scene questions in real-time with low-latency voice.
Amazon has developed a prototype for a next-generation, conversational movie recommendation assistant. The system leverages their new Amazon Nova Sonic 2.0 speech-to-speech model for real-time, low-latency voice conversations and is orchestrated by the Amazon Bedrock AgentCore framework. Unlike traditional collaborative filtering models that suggest similar genres, this agentic AI aims to understand contextual needs—like a user's mood after a long day or their desire for a lighter film after a heavy drama. It synthesizes data from plot summaries, reviews, and viewing history, acting as a personal entertainment concierge.
The technical architecture is complex, designed to handle real-time interaction. A user authenticates via a web UI, and a WebSocket connection is established to a server hosted on AWS Fargate. User voice commands are sent to the Nova Sonic 2.0 model, and the Fargate container manages an agentic workflow using a Model Context Protocol (MCP) server. Amazon Bedrock AgentCore Gateway transforms AWS Lambda functions into tools for the agent, which can perform tasks like semantic search using OpenSearch and S3 Vector. The system processes two main use cases: delivering mood-based recommendations and providing instant scene analysis, like identifying an actor or summarizing plot points upon request.
- Uses Amazon Nova Sonic 2.0 for real-time, human-like voice conversations with low latency.
- Agentic workflow powered by Amazon Bedrock AgentCore and a Model Context Protocol (MCP) server.
- Goes beyond genre to recommend based on context like mood, time of day, and social setting.
Why It Matters
Shows a future where streaming platforms offer truly personalized, conversational interfaces instead of static recommendation grids.