Why 12 phrases?
The first 10 phrases exercise every phoneme in the English language, and they’re pretty easy for most people to say. The last 2 phrases are “control phrases” used to test the AI model for effectiveness.
How will you know if it is successful?
If I can create a custom model for a person that allows them to be understood twice as much as they were before, I think that will add enough value to consider it a success.
How will the app work?
After the user installs the app, it will ask them to train the model one time speaking the 10 four-word phrases. Once the model is trained, the user will open the app, speak into it, and then the app will speak clear English to the listener.