We could theoretically have incoming audio flow through the tree graph in the same way as output is handled now, but that means the whole path of parent voices down to the root voice need to both actually receive input, and run their subvoices inlined. While this might be useful in some advanced scenarios, it's probably not what's needed in your average application.
A much simpler approach would be to add a unit "input" or similar, which directly grabs audio from the master audio interface that the root voice outputs to. All "input" units would see the same input normally. To regain probably the most interesting feature of the "proper" design outlined above, we might want some way of inserting a master input voice somewhere, that gets to process input audio before the normal graph processing begins.
We could theoretically have incoming audio flow through the tree graph in the same way as output is handled now, but that means the whole path of parent voices down to the root voice need to both actually receive input, and run their subvoices inlined. While this might be useful in some advanced scenarios, it's probably not what's needed in your average application.
A much simpler approach would be to add a unit "input" or similar, which directly grabs audio from the master audio interface that the root voice outputs to. All "input" units would see the same input normally. To regain probably the most interesting feature of the "proper" design outlined above, we might want some way of inserting a master input voice somewhere, that gets to process input audio before the normal graph processing begins.