Bringing my LED display to life with GPT-Live-1 and Codex | OpenAI Developers

This post details the process of building a voice-controlled LED display prototype by integrating GPT-Live-1 (full-duplex voice model) and Codex with a Raspberry Pi and a HUB75 LED panel. It describes how Codex assisted in hardware identification, configuration (WLED-MM), and troubleshooting for the LED panel. The project evolved from a Mac-based prototype to a self-contained Raspberry Pi system, separating voice services (microphone, wake word, GPT-Live-1 connection) from rendering services (idle screen, data refresh, frame streaming). The integration with GPT-5.6 Luna via the Responses API enabled web search, calendar access, and image lookup for dynamic display content. Key technical challenges addressed include fitting information onto a small display, moving the application off a development machine, handling conversational context for delegation, and resizing/attributing images for the display.