sound.Generate now accepts table of samples, or direct binary data as a string The table input should have float values from -1 to 1 (index starting with 1), and is about 4 times faster than using a function string input must be raw 16bit sound data (each sample is a signed short), and is much faster than the 2 other methods. EmitSound handles registered sounds a bit better clientside You could, and still can add `.wav` to the end of the registered sounds to make them work with EmitSound, which is still useful for serverside, but clientside this is no longer necessary. Added optional argument to sound.Generate - loopStart (sample ID)