a = a.astype(numpy.float16) should convert the array over for you. Note: float16 did not exist prior to numpy1.6. The default version of numpy that I got from the ubuntu repo was only version 1.5.1.
On Mon, May 21, 2012 at 1:29 PM, Mike Lawrence <mike.lw...@gmail.com> wrote: > Thanks for your response. Support for your suggestion comes from > observation that changing the size from -16 to -8 yields an octupling > of the duration. So it seems that pygame.sndarray.make_sound isn't > properly accounting for the bit depth of the numpy array. > > On Mon, May 21, 2012 at 3:03 PM, Nicholas Seward > <nicholas.sew...@gmail.com> wrote: >> Could it be that a is an array of 64bit floats and each array element >> becomes 4 16-bit samples? >> >> On Mon, May 21, 2012 at 12:54 PM, Mike Lawrence <mike.lw...@gmail.com> wrote: >>> The following code: >>> >>> import pygame, numpy >>> pygame.mixer.pre_init(frequency=96000,size=-16,channels=1) >>> pygame.init() >>> a = numpy.random.randn(96000) >>> sound = pygame.sndarray.make_sound(a) >>> print sound.get_length() >>> >>> yields a print-out of 4.0, suggesting that the specified duration of >>> 96000 samples at a 96000kHz sampling rate was somehow quadrupled >>> somewhere along the way. Any idea what I'm missing here? Or is this a >>> bug?
