Mastra adds Inworld real-time voice
Mastra has introduced support for real-time Inworld speech-to-speech capabilities via the @mastra/voice-inworld package. Utilizing WebSockets for bidirectional audio streaming, the integration inherits from MastraVoice to simplify building voice-capable TypeScript agents.
Real-time audio streaming via WebSockets represents a major leap over sluggish turn-based voice APIs, and Mastra's native Inworld adapter gives TypeScript developers a modular way to deploy voice-capable agents.
* Low-Latency Bidirectional Streaming: Uses WebSockets to run concurrent audio streaming, allowing natural user-agent conversations.
* Unified Interface: Employs the `MastraVoice` base class structure, making it simple to swap between voice providers (e.g. ElevenLabs, OpenAI) with minimal code changes.
* Native Orchestration: Connects directly with Mastra's existing agent framework, linking memory, tools, and workflows to the voice channel.
DISCOVERED
1h ago
2026-06-10
PUBLISHED
1h ago
2026-06-10
RELEVANCE
AUTHOR
mastra