The title of the thread should say it all....
Which lossless CODEC will give the highest compression ratio from the original?
I was just looking at a thread about MP4 Audio lossless which achieves about 1.4:1. Are there any CODECs which will achieve anything like 4:1 in lossless?

depends on the signal. silence compresses really good

Impossible for most audio you're going to compress. If you want 4:1 - use OGG ..

http://wiki.hydrogenaudio.org/index.php?ti...less_comparison
links for comparison on the bottom.

1:4 is possible with all compressor when the signal is highly compressible; it's impossible in most situation.

My tests show that among commonly used encoder the best compression ratio is offered by Monkey Audio in Extra mode...

...and what kind of ratio does that offer?

That was already answered in this very same thread. Look pest's post above: depends on the signal.
Best AVERAGE ratio I've seen is with LA, but it has its drawbacks....

FLAC compressed the first Harry Potter audiobook at 5:1.

OK, thanks for that.

That 5:1 ratio was almost certainly for a mono source. When I do mono LP transfers, I make the final product truly mono -- identical left and right channels. I back up my work with MA. Making the two channels identical, rather than using the 'almost the same' results of normal processing, produces markedly smaller compressed backup files.

My experience too. Best to me among the 'fast enough encoding and decoding' codecs.
I don't think we will ever get a compression ratio seriously better on average.

The one which offers the best compression possible (yes, even better than LA) is OptimFROG, it's also slowest.

Nope. WavPack -hx6 is even slower. Although I haven't tested ofr --maximum-compression yet.

In my test corpus (results soon to be posted), I note a very strong competition, ratio-wise, between LA -high and ofr --mode bestnew --seek slow. Not even WavPack can come close to their compression. But LA always comes out faster than ofr.

Nothing beats the speed of FLAC though.

Note: Also haven't tested the yet-to-be-named lossless compressor by TBeck.

For what? Encoding: WavPack -f does. Decoding? Shorten does it too

Something like "--optimize best --mode ultranew --seek min --speedup 1x" should give most compression, and I doubt that you'll find anything slower.

@rutra80: Yes that's --maximum-compression as Ghido has explained.

@guruboolez: But does anyone actually use WavPack -f ? I doubt it. Most of use will use WavPack -h or even -hx or (God forbid) -hx6. And I don't think Shorten is that popular anymore... FLAC has replaced it for many purposes.

wavpack -f is faster than FLAC for decoding too, according to your own test :B



Guruboolez himself (not that I have any clue why)

You're wrong, --maximum compression = --optimize best --mode ultranew --seek normal --speedup 4x.

OK, now I'm lost!
Let's say I don't care about how long it takes, but that I'm just interested in conversion of stereo 44k 16bit wav file (music) into a lossless format with the highest reduction in disc space consumption?

Try OptimFROG with the settings I provided in 15th post. Note that you won't be able to play these files in real-time (unless you have 4GHz CPU or something like that).

Right, but if WavPack -f appeared as faster on the decoding side, it was only by a very slight margin. Moreover, I experienced the opposite with foobar2000 0.9 (again, the difference in favor of FLAC is really small - and to increase the confusion, these results are varying from file to file: some are in favor of WavPack, and others aren't... ).
But on my laptop, flac is significantly faster on decoding than WavPack (x150 vs x120). Now if people will ask my why I didn't performed my (old) test on my laptop, it's simply that my RAM was too short at this time (256 MB with XPSP2), making a complete buffering impossible. Since I bought 512 additionnal MB
Therefore, I wouldn't say anymore that WavPack is faster: speed measurement varies depending on file (only a little) and much more on CPU.


N.B.
desktop: AMD Duron 800
laptop: AMD Athlon



audio2u> use OptimFROG with --maximum-compression if you want the best encoding ratio and nothing else. You should nevertheless take note that nobody on this board (excepted maybe one or two persons) is using the strongest formats (LA or Frog) as shown in the latest poll. It's not that people aren't concerned by efficiency: it's simply that using such strong encoders isn't really friendly for daily usage (playback, reencoding, burning, (im)possible hardware support, etc...)


EDIT: I always used -fx5 with WavPack (I left this setting once for hybrid, but I quickly changed my mind and used -fx5 back).

LA generally offers the best compression ratios, but is poorly developed, and buggy at times. generally, OptimFROG achieves the next highest ratios. imho, monkey's audio and wavpack also can achieve very good compression ratios, and may be worth considering, but if compresison ratio really is your only criteria, LA or OptimFROG are probably your best bets.

I'm not sure that LA compresses better than OptimFROG. All comparison concluding on LA's superiority (Heijden's one and Speek's one) aren't using OptimFROG with --bestnew but --highnew only (though LA -high and Frog --highnew are much more comparable in term of speed).

Johann de Bock's and my own comparison are based on the highest Frog setting (not mine anymore though) and both are showing that LA produces slightly bigger files. It might be interesting to note that MPEG-4 ALS has coding tools allowing ultra-high compression too which are possibly more efficient (not for speed but for space reduction) than LA. MPEG-4 ALS is possibly the second best compressor at the moment.

In that case, you'd best use a format that's not meant for playback, but pure archiving. Otherwise, just pick your favorite format.. Compromises are good, but generally, you're either aiming for speed or space, not both.

Since OP wants space, so LA -high then. Don't forget to store the la.exe file somewhere save... I'm not sure it will last...

OTOH, I have a file which ofr compresses very nicely: 28,461,955 for the .ofr file vs 33,316,684 for LA. Decoded .wav is binary identical.

So then I guess ofr --maximumcompression is also okay. Still playable realtime with a 2GHz processor.

Edit: One caution, LA does not seem to support tagging.

Aha. TTA is now the undisputed king of encoding speed. Faster than WavPack -f in 100% of the files I tested up to now.

tar is even faster

Read this:
http://www.hydrogenaudio.org/forums/index....ndpost&p=386514

TTA performance are just a copy of Monkey's -fast mode. It's anything but fast (excepted maybe on the encoding side, but even here it's still clearly not a king).

OptimFROG's --maximumcompression is exactly equivalent to --mode ultranew --seek slow --optimize best.

I STRONGLY recommend NOT to use the deprecated --speedup option. There have been some posts which say that using --speedup 1x obtains better compression. As a courtesy for you and to end any doubts, I tested it on my reduced corpus (80 one minute fragments from Audio-CDs / 846 MB), and the results were very clear.

Comparing --maximumcompression and --maximumcompression --speedup --1x, I obtained the following:

- compression median +0.01455% (--speedup 1x is worse)
- compression average +0.00664% (--speedup 1x is worse)
- 78 of 80 files have a difference between them between -0.05% and +0.05% (very small)

Aside the fact that --speedup 1x several times slower, it is actually worse on average. Please believe me, because as the author of OptimFROG, I can assure you that --speedup 1x cannot be better on average.

If you really want to experiment, you can also try --maximumcompression --seek min (or equivalently --mode ultranew --seek min --optimize best), and select the best of the two. But, if you do not select the best by hand, you should stick to using --maximumcompression only.

This --maximumcompression switch was added especially to avoid using long switches and confusion for achieve the MAXIMUM compression.

OptimFROG has also very fast modes. For example, WavPack at its highest level (wavpack -h) decodes at 50x, and OptimFROG at the FASTEST level (ofr --mode fast) decodes at 34x, already obtaining 0.63% better compression (result obtained on a 80 CD / 50 GB audio corpus).

The mode ofr --mode bestnew --seek slow --optimize best already obtain 0.2% better compression than la --high (on the same 50 GB audio corpus). Additionally, LA fails badly on a significant number of files, with compression worse with 2-7% than OptimFROG.

OptimFROG highest setting is clearly better than LA. The --maximumcompression works well for realtime decoding. It uses 60% of CPU for realtime decoding on my 2.8 GHz P4.

BTW, there is available (at http://www.LosslessAudio.org/) since yesterday a new beta version (4.520b1) of OptimFROG. The new promised experimental version, with significant improvements in compression (0.6% on average), is almost ready and probably will be released next week .



Thanks to all for your support,
Florin Ghido