When working through the Speech Synthesis post last night, I kept getting a nagging feeling that the speech code I was putting in my ViewModel just didn’t really belong there. The speech was the UI, it was the thing the user actually interacted with, so it smelled wrong to have it in the ViewModel. It felt like I was putting textboxes and button-emitting code inside VM functions.

That said, Speech isn’t exactly a traditional visual View it’s a …uh…sound :)

I’m looking at building some more samples that have deeper speech integration, including recognition, but I don’t think that code should be in the ViewModel. At the same time, it represents a significant amount of functionality that in a real app would need to be tested, and like a View, it doesn’t test overly well programmatically.

So, MVVM / ViewModel patterns Gurus and moderately interested others: where would you put this code?

Would you create a separate class that interacts with the viewmodel just like the View does, parsing input and then calling VM functions, and call it an “Audible View” or something? What would own it, and how would you instantiate it? Would you tie it to a visible View?

Would you just put the code in the ViewModel and say the “real” UI is the view, and the speech is just ancillary? That could make for a pretty hefty viewmodel, with lots of things that aren’t really testable.

Would you do something else?

Help me readers, you’re my only hope.