[MA35D1] Implementing Real-time Voice-based ChatGPT

This example code demonstrates how to build a real-time, voice-interactive conversation system on the Nuvoton OpenWrt embedded platform using Python and WebSocket technologies, integrated with OpenAI’s latest Realtime API. After a user speaks into a microphone, the device records and uploads the audio in real time. The OpenAI model then quickly generates a voice response, which is immediately played back by the system, making human-computer interaction more natural, fast, and intuitive.

This low-latency interaction capability enables embedded platforms to support applications such as intelligent voice assistants, voice-based customer service, and educational chatbots. Even on resource-constrained devices, leveraging cloud-based AI allows for a voice interaction experience comparable to that of high-end devices like smartphones.

This example code is developed based on the Nuvoton OpenWrt 22.03 environment. You must first set up the Nuvoton OpenWrt 22.03 development environment (https://github.com/OpenNuvoton/Nuvoton-OpenWrt-22.03) before running this example.

While this document uses the NuMaker-IoT-MA35D16F90 evaluation board as an example, the content applies to any evaluation board that supports Nuvoton OpenWrt 22.03 and can run Python smoothly, including the MA35D0, MA35D1, and MA35H0 series.

You can download the sample code at EC_MA35D1_Real-time_Voice-based_ChatGPT_V1.00