Earth Mountain almost there, page-35

It's my own speculation, not really 'info' but based on my observations in the earlier youtube video you could see the chip was being driven from a MATLAB environment and a large donut-shaped circuit board which had an FPGA chip. To play sound, the desired sound was selected in MATLAB and the pre-processed to convert it into the required commands/data for the audio pixel chip. For a finished product, the conversion would need to occur in real time to stream commands from a digital source to one or more audio pixel chips.

There was some talk in an announcement last year about the software being 'adapted' for this purpose - which makes sense to me. They would have used the familiar MATLAB environment with its extensive debugging and investigative tools to develop the algorithms (and it appeared, they tried lots of different algorithms presumably to see what worked best for different sounds) and then at the end, the selected algorithms will need to be re-written in efficient, optimised code suitable for use in a real-time streaming application. This code would presumably run on either an embedded chip, or perhaps involve an FPGA or other hardware - this is something announcements have been relatively quiet on, but it doesn't seem like a showstopper... if they can get the chips to produce sound, I think they can figure the software and off-the-shelf chips out to support it.