This seems to be begging for an FPGA Tensorflow back-end. To my knowledge, quality speech-to-text is quite compute intensive; and can't run with reasonable late

This seems to be begging for an FPGA Tensorflow back-end. To my knowledge, quality speech-to-text is quite compute intensive; and can't run with reasonable latency on a Raspberry Pi Zero, for example. Coupled with a chat-bot you could have a smart anything. "Laundry machine, start the last program" "Dang, Sarah, what did you do in these clothes? Smells really weird. Starting nuclear bio-hazard cleansing." "Steeeeve, imma kill you! Why you always have to switch it to the smart-ass mode?" Yikes