在 AWS 上构建超低延迟语音 AI 应用
原标题：Building ultra-low-latency voice AI applications on AWS
AI 摘要
AWS 机器学习博客介绍 AWS 合作伙伴 Cartesia 如何利用状态空间模型（SSM）和 AWS 基础设施构建超低延迟语音 AI 应用。Cartesia 推出 Sonic（文本转语音）和 Ink（语音转文字）两个模型，均实现低于 100 毫秒的首字节延迟，其中 Sonic 约 90 毫秒即可输出首个音频字节。文章对比了 Transformer 在流式延迟上的架构局限，指出 SSM 具备恒定内存占用和原生流式支持，更适合实时语音场景。
正文节选
Building ultra-low-latency voice AI applications on AWS Conversational AI is transforming how humans interact with technology. From virtual assistants and AI-powered contact centers to real-time translation and accessibility tools, voice-first interfaces are rapidly becoming the default mode of interaction for millions of users worldwide. Yet delivering a truly natural voice experience remains one of the hardest challenges in AI infrastructure. Users expect human-like responsiveness. Excess late