Lossless Audio Codec Compression Comparison

[mg262]

I need to archive the audio from some captured VHS tapes, so I ran a quick comparison of some lossless codecs. The source was about 25 minutes of mono 16-bit 48,000 kHz unprocessed WAV PCM audio from an old children's cartoon; options were set to maximise compression.

Program Options Size Compression
OptimFROG 4.509 ofr --encode --mode bestnew 73,031,530 51.16
LA 0.4 la -high -noseek 73,080,442 51.20
OptimFROG 4.509 ofr --encode 73,518,268 51.51
Monkey's Audio 3.99 Extra High* 73,679,676 51.62
WAVPACK 4.3 wavpack -h 74,625,642 52.28
LPAC 1.40 lpac -5 74,751,176 52.37
TTA1 3.3 ttaenc -e 75,036,829 52.57
flac 1.1.2 flac -8 75,313,850 52.76

[Original] 142,740,044 100.00*Insane performed worse.

I was surprised at how close all the figures were, especially when compared to lossless video codecs.

(By the way, does anyone know a way to frameserve audio data from AVIs to a compressor rather than having to extract to a WAV?)

[AMTuring]

Did you keep track of the time to compress also?

[Kurtnoise]

options were set to maximise compression.
in this case, use bestnew switch for optimfrog and have a look at Monkey's Audio and LA codecs...

I was surprised at how close all the figures were, especially when compared to lossless video codecs.
huh ? could you develop a little bit more please...

(By the way, does anyone know a way to frameserve audio data from AVIs to a compressor rather than having to extract to a WAV?)
BeSweet can do that...it depends of course of your audio stream included in the AVI.

[Rockaria]

(By the way, does anyone know a way to frameserve audio data from AVIs to a compressor rather than having to extract to a WAV?)
That's what the avisynth can do as is introduced here. (http://forum.doom9.org/showthread.php?t=103069)

a = AviSource("D:\captures\jewel.avi") will read the a/v frame to the clip a, the audio stream of which can be input to any pipe-in capable encoder with the help of avs2wav or bepipe.

[mg262]

@AMTuring,
No, I didn't, I'm afraid. But IMO looking at compression and speed would need many tests/codec (with different option sets).

@Kurtnoise13,
Thank you for the suggestions; I've updated the table.

I was surprised that the various codecs available compressed by almost exactly the same amount -- the variation between the best and worst was only 1.6%. Cf. a lossless video codecs comparison (http://forum.doom9.org/showthread.php?p=610995#post610995), where the difference between best and worst is larger. In particular, with video codecs there's a real time/compression tradeoff (although that comparison doesn't show it very well): spending more time can give significantly better compression (e.g. FFV1 and VBLE). With audio, very fast codecs (e.g. tta) do almost as well as very slow ones.

Thank you for the BeSweet pointer... I'll look into it.