Skip to content

Latest commit

 

History

History
55 lines (43 loc) · 2.54 KB

ffmpeg.md

File metadata and controls

55 lines (43 loc) · 2.54 KB
Raw

ffmpeg

ffmpeg appears to be actually a general-purpose audio/video mixer-synthesizer (it includes a sine wave generator, for example, as well as, for example, fade-in/fade-out effects); this does not seem to be mentioned very much, if at all, probably because its documentation is very difficult to understand.

After a number of false starts, I restarted with this segment of a level-two debug output:

1631591999.73#  DEBUG: schedule: [0 [(play) (72699__benboncan__cuckoo) 0.05 10.65] 64]
1631591999.73#  DEBUG: schedule: [0 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [1.71428571428571 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [3.42857142857143 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [5.14285714285714 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [6.85714285714286 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [8.57142857142857 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [10.2857142857143 [(play) (72699__benboncan__cuckoo) 0.05 10.65] 64]
1631591999.73#  DEBUG: schedule: [10.2857142857143 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [12 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [13.7142857142857 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [15.4285714285714 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [17.1428571428571 (noteon 1 48 64)]
1631591999.73#  DEBUG: schedule: [18.8571428571429 (noteon 1 48 64)]

Ignoring the MIDI notes, I tried to construct an ffmpeg filter that would merge the cuckoo recording with a delayed copy of itself, and (after a lot of trial-and-error) ended up with the following, which worked:

cuckoo=data/72699__benboncan__cuckoo.wav
ffmpeg \
    -i "$cuckoo" \
    -i "$cuckoo" \
    -filter_complex '[0]atrim=0.05:10.65 [g0]; [1]atrim=0.05:10.65,adelay=10286 [g1]; [g0][g1] amix=inputs=2'  test.oga

Volume adjustments

Volume adjustments in ffmpeg can be written as fractions, e.g., 64/127, which is great considering MIDI velocities range from 0 to 127. However, volume adjustments appear to be global, so

    [0]atrim=volume=64/127,0.05:10.65 [g0]; [1]atrim=volume=64/127,0.05:10.65,adelay=10286 [g1]; [g0][g1] amix=inputs=2

looks correct but produces incorrect results. After the volume is turned down for input 0, the volume is turned down again for input 1.

So apparently, this has to be written as

    [0]atrim=volume=64/127,0.05:10.65 [g0]; [1]atrim=0.05:10.65,adelay=10286 [g1]; [g0][g1] amix=inputs=2

which looks incorrect but actually produces correct results.