This project aims to research google's offline speech recognition, from several android apps and ideally make them interoperable by replicating it on any system that supports tensorflow.
Why do you think that https://github.com/espressif/esp-adf is a good alternative to google-offline-speech-recognition