I went the full way with googles “Micro Speech Training” including porting to ESP32 and getting the python training part set-up + running. Its a nightmare.
Thanks to your solution, its exactly taking away this complexity at the right places - and still providing the flexibility to modify. Within one evening I get the same result - compared to many hours of work. COOL
ML on EdgeDevices is key technology, I am working in Pharma and SupplyChain - and it has great capabilities.