Architects and implements high-performance, low-latency voice agents for natural human-AI interaction.
This skill equips Claude with the expertise of a production-grade voice AI architect, focusing on minimizing latency to achieve conversational flows under 800ms. It provides patterns for both Speech-to-Speech (S2S) models and modular STT→LLM→TTS pipelines, while addressing critical real-world challenges like voice activity detection (VAD), barge-in handling, and turn-taking logic to ensure fluid, human-like dialogue across various voice interfaces.
