-
There are actually 32 voices - the top 16 are samples, the bottom 16 are the live input. If there is no sample loaded for a voice (for the top 16), or there is no live input (for the bottom 16), the weighting pixel will be red, so changing the weighting of red voices in the grid will have no effect (since there is nothing there).
-
V/oct allows controlling of the pitch of grains - like freeze and reverse, it is polyphonic - the first input controls the grain pitches of the 16 sample voices, the 2nd controls the grain pitches of the 16 live inputs
-
The Freeze and Reverse inputs are polyphonic. Since polyphony is limited to 16 voices, it takes 2 inputs to control the freeze and reverse (since each voice can be frozen or reversed individually). The push button switches between gate mode (where a high signal freezes/reverses the voice) and trigger mode where each low to high trigger changes the freeze/reverse state.
-
Scaling chooses between a linear weighting to an exponential weighting of the grain choosing probability. (its pretty subtle, especially with only a few grain voices)
Hope that helps