ESP-BOX ChatGPT demo

What you’re building

Espressif’s ChatGPT demo uses an ESP32-S3-BOX, BOX-Lite, or BOX-3 as a voice chatbot terminal. The device records spoken input, sends it through OpenAI APIs, receives a response, converts the response to speech, and plays it through the built-in audio hardware while using the screen for interaction. ESP32-S3-BOX-3 is explicitly supported. The repository requires ESP-IDF master, with a specific fallback commit documented for build problems, and contains two projects that must both be built: factory_nvs for stored credentials and chatgpt_demo for the application. The expected result is cloud-backed voice conversation after entering Wi-Fi and API credentials through on-screen prompts. This is not offline inference: it requires internet access, an OpenAI API key, and available paid API tokens. The source also says the example is unavailable in Mainland China because of OpenAI restrictions.

Key steps

1. Install ESP-IDF master, or use the README’s pinned fallback commit if component builds fail, and update submodules recursively.
2. Clone Espressif’s esp-box repository and enter examples/chatgpt_demo/factory_nvs.
3. Run idf.py menuconfig, select ESP32-S3-BOX-3 and the appropriate HMI board configuration, then build factory_nvs.
4. Move to examples/chatgpt_demo, repeat the BOX-3 hardware selection, and build the main demo.
5. Flash the main project to the connected serial port and let the board boot.
6. Follow the display prompts to enter Wi-Fi credentials and a valid OpenAI key, then test spoken input and audio response.