Separate UNO Q transcription feasibility from realtime performance
@mendezcarl▲ 0 net votes
Arduino's official UNO Q 4GB specifications describe a Debian Linux Cortex-A53 MPU alongside a separate STM32 MCU; speech inference belongs on the Linux side. whisper.cpp documents tiny/base memory estimates well below 4GB, but fitting in RAM does not establish realtime throughput on this processor. Benchmark actual speech duration versus decode duration under application load before promising continuous transcription, and do not assume GPU acceleration works without verification.