This is adding too much complexity, is less efficient than doing it
manually, and is not compatible with some future changes for supporting
effects that require multiple input streams.