AssemblyAI Launches Universal 3.5 Pro Realtime Integration for LiveKit
AssemblyAI has released an integration for its Universal 3.5 Pro Realtime speech-to-text model into the LiveKit agents framework, enabling developers to build voice agents with live conversation context, voice focus, and server-side turn detection. The model is specifically designed for real-time applications and supports multilingual capabilities and programmable turn-detection parameters. This integration facilitates low-latency speech processing for LiveKit-based voice projects.
AssemblyAI Launches Sync API for Low-Latency Short-Form Transcription
AssemblyAI launched the Sync API, an HTTP-based service for transcribing short audio clips (up to 2 minutes). By sending an audio file in a single request, users receive a transcript in approximately 134 ms, eliminating the need for polling or persistent WebSocket connections. The API utilizes the Universal-3.5 Pro model, offering high-accuracy transcription for dictation, voice agents, and IVR workflows without the latency associated with traditional asynchronous batch processing or the overhead of real-time streaming sessions.
AssemblyAI resolved an issue in its dashboard that caused errors when updating payment methods. A change in the Stripe response schema had resulted in 400 errors for users attempting to modify their payment details; this service has been fully restored.
