Hi!

cPatPmtGenerator::MakeLanguageDescriptor in remux.c incorrectly flags all audio tracks as "clean effects", which means a track without any voice or language.

The proper solution would be to preserve the audio_type of the original stream (which may indicate e.g. a visual-impaired audio track, as used for e.g. automated subtitles-to-speech audio tracks by YLE here in Finland).

The attached patch is a simpler stop-gap measure that simply always sets the audio_type to the value of 0x00 which means a "normal" track, which is obviously better than marking the track as a special "clean effects" one.