This repo contains a Dockerfile and a couple of commands to work with Deepspeech v0.7
You can use record.py script to record audio samples to train a model for specific commands. Records will be stored in training-data/dev, test and train directories