That's mostly correct, as I posted in the other thread, the NuVoice N570S64A chip used in these decoders has 8 MB (64 Mbits) of built in flash memory used to store the sound and SPI (Serial Peripheral Interface) to load it. The "Serial" part of that is the key, it's one bit at a time, so it's fairly slow.
You can go to the sticky thread at the top of this forum and actually load and play the spirom files, the loading in the browser takes somewhere around 20-30 seconds or so, much faster than the load times into the locos.