📄 中文摘要
开源语音模型正在迅速增加，但在设备间缺乏统一的本地推理平台来支持语音代理工作负载，包括转录、实时流媒体、说话人识别、语音活动检测和实时翻译等。ExecuTorch 提供了一个跨平台的解决方案，旨在简化这些任务的实现，使开发者能够在不同设备上高效地构建和部署语音代理。该平台不仅支持多种音频处理功能，还优化了性能，确保在各种硬件上都能实现高效的语音识别和处理。通过 ExecuTorch，开发者能够更容易地创建智能语音应用，推动语音技术的普及和应用。
使用 ExecuTorch 构建语音代理：跨平台的设备音频基础
📄 English Summary
Building Voice Agents with ExecuTorch: A Cross-Platform Foundation for On-Device Audio
Open source voice models are rapidly proliferating, yet there is a lack of a unified native inference platform for voice agent workloads across devices, including transcription, real-time streaming, diarization, voice activity detection, and live translation. ExecuTorch offers a cross-platform solution aimed at simplifying the implementation of these tasks, enabling developers to efficiently build and deploy voice agents on various devices. The platform supports a range of audio processing functionalities and optimizes performance to ensure efficient speech recognition and processing across different hardware. With ExecuTorch, developers can more easily create intelligent voice applications, promoting the widespread adoption and application of voice technology.