Normalize and compress audio for oral histories

[eddierubeiz]

Ref #3215

Targets:

• Integrated loudness (I): -19 LUFS
• True peak (TP): −1.5 dBTP
• Loudness range (LRA): ~7 LU

We're also making ffmpeg do a bit of audio compression, ensuring that the loud and soft parts within the file are brought closer to each other (within roughly 7 loudness units), in addition to the entire file being raised / lowered to about -16 LUFS.

Results

ffmpeg -i combined_audio.m4a -filter:a ebur128=framelog=verbose -f null -

Before:

Integrated loudness:
I:         -23.1 LUFS
Threshold: -34.4 LUFS

Loudness range:
LRA:        12.5 LU
Threshold: -44.4 LUFS
LRA low:   -32.3 LUFS
LRA high:  -19.7 LUFS

After:

Integrated loudness:
  I:         -21.9 LUFS
  Threshold: -32.6 LUFS

Loudness range:
  LRA:         7.2 LU
  Threshold: -42.6 LUFS
  LRA low:   -26.5 LUFS
  LRA high:  -19.3 LUFS