Defining moment of the audio attenuation through ffmpeg

Question

There are audio tracks of different lengths in m4a format. And there's ffmpeg library for working with the media. Many of the tracks have the effect of "decay" in the end, and it is necessary to determine at what point it occurs (determined once and the value entered in the database along with other information about the track). Those. we must somehow determine that the track begins to fade, and its volume reached 30% compared to the total volume of the song. Is it possible to solve by means of ffmpeg, and if so, how?

Answer 1

If you run this command,

ffmpeg -i in.mp4
      -af astats=metadata=1:reset=1,
          ametadata=print:key=lavfi.astats.Overall.RMS_level:file=vol.log -vn -f null -

it will generate a file called vol.log which looks like this

frame:8941 pts:9155584 pts_time:190.741
lavfi.astats.Overall.RMS_level=-79.715762
frame:8942 pts:9156608 pts_time:190.763
lavfi.astats.Overall.RMS_level=-83.973798
frame:8943 pts:9157632 pts_time:190.784
lavfi.astats.Overall.RMS_level=-90.068668
frame:8944 pts:9158656 pts_time:190.805
lavfi.astats.Overall.RMS_level=-97.745197
frame:8945 pts:9159680 pts_time:190.827
lavfi.astats.Overall.RMS_level=-125.611266
frame:8946 pts:9160704 pts_time:190.848
lavfi.astats.Overall.RMS_level=-inf
frame:8947 pts:9161728 pts_time:190.869
lavfi.astats.Overall.RMS_level=-inf

The pts_time is the time index and the RMS level is the mean volume of that interval (21 ms here). Each drop of 6dB corresponds to a drop of half the present volume.

If you run the command with reset=0, the last reading in the generated log file will show the RMS volume for the whole file. Then the volume which is 30% of the mean volume is ~10.5 dB below the mean value.