I wonder if these models can be used for speech impairments related to other neurological diseases like Parkinson's which seems to cause issues with speech clarity as well as loudness.
Also, in other areas like NLP, Google has released tools to train your own models. I wish they would do that here as well. They could potentially get access to a lot more training samples if people started training their own models and testing the performance. The benefit to people suffering from speech impairments could be immense if such a technology were more accessible.