Closed marek-cottingham closed 3 years ago
Changing the synthesis frame to not a constant frameSize//16 was rejected, on the basis that this leads stretches at higher factors, x20, x24, x32 to take an extremely long time to compute. There is no notable change in audio quality for a stretch of x8 between the two options.
The size of the appropriate synthesis hop relative to the frame width may be proportional to stretching factor. Test if this is the the case with x8, x24 stretches and if so, hard code it into the methods (DataSet_1D).