This is a COSGC wearables project that runs speech recognition entirely on an ESP32-S3, with no cloud connection. It is based on Oído, an open on-device speech-to-text project by Lokutor, which runs English and Spanish models on a roughly $5 chip.
My part is the hardware build: a PlatformIO project for an ESP32-S3 N16R8 board with an INMP441 I2S microphone and an SSD1306 OLED display, plus the wiring notes to go with it. The speech models and most of the repository’s code and research come from the original Oído authors.