This ASR model uses Vosk for ASR. It is rather easy to use. You can put specific words in the vocabulary.txt file to increase the accuracy of the model.
You can then import it and use the listen function to start ASR.
there's also a speak function to do offline TTS.