I use speech-to-text on my iPhone inside a gmail draft notes on the go as well. I'll keep adding to the draft then email it to myself. I'll also use VoiceRecorder app that can do speech-to-text after the fact.
Can't agree more with your GCP vs AWS mention. I have deployed my share of models on both platforms and GCP, with Cloud ML is much much easier esp. with Tensorflow. That said, hoping that 2019 makes it even easier as its still too complex for my taste.