This is a sample APK that connects to Termux for use in transcription, as well as a demo of offline speech recognition. The last release wasn't properly signed, so I've fixed that here. I've also included a demo Termux script for an example of how to interact with the app (using a .wav file in the Music folder of your device). Read below for more information on how it works:
This is a simple app made from the core components of Athena I was using for another project called Spellbook. The key useful feature here is that it takes an audio file and transcribes the text into a timestamped file in the documents folder. You MUST give it access to your documents folder for it to work. After the transcription it sends a command to run a file titled "sapphire-core.py" in the Termux home directory, with the transcribed file name as a command line argument. This was all designed for me to implement prototype features for the sapphire framework on my phone using whichever languages fit best before rewriting them for Android natively. The transcription can be started by sending a intent from Termux to Spellbook, and therefore used to build toy voice controlled scripts for now. You can find that sample Termux command in the assets of the apk.
I'll be building on this more and making it easier to use, but I figure it's long overdue to put out some kind of usable apk for people. You can also use this to demo general voice dictation and wake word recognition (default wake word is "computer")