Ignoring descriptive subtitles

First of all, thanks a lot, this is a great tool and now I can watch my new favourite series without being spoiled by the subtitles :-)
In the description you say that sometimes the timing is not perfect, and in the case of my series I think I know why: there are a lot of descriptive subtitles like [COUNTRY MUSIC PLAYING] or [METALLIC NOISE], and the algorithm also consider those subtitles to align them with speech whereas no one is speaking. So I guess if in the parsing we could ignore these kind of subtitles, usually in capital letter and between brackets, then the alignment would be better. The discarded subtitles could be either completely discarded or put back between their 2 surrounding normal subtitles, optionally with a linear interpolation to put them closer to the subtitle they where originally closer to.
Again, thanks a lot!

The issue does not name files or tests. Start by locating the subtitle parsing and alignment entry points, then trace how descriptive lines such as bracketed, uppercase sound cues are handled. Done should mean these cues no longer distort speech alignment, with their handling matching the chosen behavior.