|
Just been looking into some AC3 time-stretching using ffmpeg/sox/aften for the decode/stretch/encode steps (basically to shorten a movie's AC3 soundtrack by 160ms).
Just wondering if there are any hidden problems in using this toolchain/workflow.
I notice that the output .ac3 has the dialnorm gone (according to eac3to), presumably because when ffmpeg transforms the input .ac3 to multichannel .wav, it gets 'lost' (discarded?).
Not sure if ffmpeg is ignoring the dialnorm in doing this conversion, or if it would be worth a preliminary .ac3->.ac3 pass with eac3to to remove dialnorm?? What do those in the know think? Would a preliminary pass (.ac3->.ac3) with eac3to have any other side-effects (don't know too much about bit-depth or any other factors - would anything else be adjusted?)
Cheers in advance,
m
|