Sam Sharp's `Sam.Speak` and `Sam.SpeakPhonetic` return a `byte[]`. For Samity we currently create a new `float[]` and convert every value. Investigate constructing the final output as a float, or working entirely with floats during audio synthesis.