Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Lossless codec comparison - part 2: hi-res (Apr '22) (Read 637 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Lossless codec comparison - part 2: hi-res (Apr '22)

Following comparing the performance of lossless audio codecs on multichannel material last month, here are the results of hi-res material with the same approach.

Report can be found here: audiograaf.nl/losslesstest/Lossless audio codec comparison - revision 5 - hires.html

The main result of the test (comparing sound sources with sample rates up to and including 192kHz) is inserted below:

X

For more results and an analysis/discussion refer to the report linked above.
Music: sounds arranged such that they construct feelings.

Re: Lossless codec comparison - part 2: hi-res (Apr '22)

Reply #1
When considering the graphs for individual sources, the most peculiar result is inserted below. This is sound sourced from a Game Boy Emulator, and thus I consider it chiptune. The emulator used is game-music-emu-0.6.3. I've changed the emulator to output 192kHz instead of 44.1kHz. Compared to the other sources, this is the only source with an appreciable amount of signal in supersonic ranges (i.e. sound above 20kHz), which is why I included it.

However, the results differ a lot from the average results. A few remarks:

  • Shorten almost beats Monkey's Audio with Insane preset
  • Monkey's Audio, OptimFROG, WavPack, TAK and ALS have faster presets outperform slower presets
  • WavPack outperforms OptimFROG by more than 10% (3%-point)
  • The difference in compression between the best and worst performing codec is more than 100%

X

Music: sounds arranged such that they construct feelings.

Re: Lossless codec comparison - part 2: hi-res (Apr '22)

Reply #2
Good work.

WavPack -hx4 costs CPU, but compared to Monkey's Insane you have regained it when you played it twice.
Which is a bit of an artificial comparison when Insane doesn't perform better.

Last two months' worth of foobar2000.org ad revenue has been donated to support war refugees from Ukraine: https://www.foobar2000.org/

Re: Lossless codec comparison - part 2: hi-res (Apr '22)

Reply #3
TAK -p0 is the fastest lossless encoder on stereo, not surprising to those who read your earlier tests.

There are only eight sources here, but TAK -p0 is fastest encoder on all -  although for the chiptune, the two encoding runs do differ so much that there is a margin of error question between TAK -p0 and FLAC -0.
TAK -p1 is second at decoding speed on most sources - and TAK -p0e and FLAC -0 trade blows quite evenly.

TAK -p1 also matches up 4--4 against Monkey's High in compression level, although one is practically tied.
Even if that is because Monkey's couldn't translate its CDDA performance on High and Insane to hires signals, it is still damn impressive.



WavPack -hx4
also decodes faster than -h. Not that it would be much noticeable in practice - except if you got days of chiptune to RG scan - on that signal, the difference is pronounced. It is likely nothing more magical than having less residual data to decompress?
Last two months' worth of foobar2000.org ad revenue has been donated to support war refugees from Ukraine: https://www.foobar2000.org/

Re: Lossless codec comparison - part 2: hi-res (Apr '22)

Reply #4
I'm usually using ffmpeg for hires flacs (if anyone's interested :-) ). Takes longer to encode but results are better, size-wise.

Re: Lossless codec comparison - part 2: hi-res (Apr '22)

Reply #5
I'm usually using ffmpeg for hires flacs (if anyone's interested :-) ). Takes longer to encode but results are better, size-wise.
ktf has made a few improvements that might one day make it into reference FLAC: https://hydrogenaud.io/index.php?topic=120158
You can download the (totally unofficial!) builds and see if you get the same improvements that I got. I got quite some differences between CDDA and high resolution.
(The last build posted was actually a bit buggy, so when that is fixed, I predict that the better compression would come cheaper. And yes the "subblock" name could be improved, I have proposed the name "combinatukey" :) )

Feel free to test and post in that thread what I suggested in #74 - and maybe the same build with
flac -8 - A partial_tukey(1);partial_tukey(2);partial_tukey(4);punchout_tukey(7);gauss(5e-3)
against ffmpeg with -compression_level 12 -lpc_type cholesky -lpc_passes 4 (or if you want subset compliance: replace the "12" by "8").
Last two months' worth of foobar2000.org ad revenue has been donated to support war refugees from Ukraine: https://www.foobar2000.org/