Show HN: Live speech-to-text on an ESP32-P4 microcontroller, no cloud
A project achieving live English speech recognition on an ESP32-P4 without cloud connectivity by running a 65.7M-parameter streaming transducer with 2/4-bit weights on the chip alone.
Key points
- •A 65.7M-parameter streaming transducer with 2/4-bit weights is run on the chip alone.
- •It operates faster than real time without requiring any cloud connection.
- •The project includes six voice-control demos.