separateAudioRenditions

True when each alternate sound arrives by a download of its own, as an HLS or a DASH audio rendition does (#455), rather than in the same bytes as the picture, as every track of a file does. Reading such a sound costs bandwidth whether anyone hears it or not, so the engine then reads only the sound being heard. A switch to another one reads it from the moment playing, by a seek of the reads that keeps what the other streams already hold when the source can seek, and the sound being heard goes on until the new one covers that moment.

False, the default, keeps every sound read into a cache, which is what makes a switch on a file instant.