Why isn't WavPack more popular?

[Mnyb]

And even more compression is not needed , it would be single digit percent better , why with today's storage.

Flac can do multichannel but not DSD , but DSD is useless anyway just convert to multichannel pcm and then pack as flac .

Here i dont know how work intensive is wavpack to uncompress ? Flac is made so that the compression is the most computer intensive part , uncompress for playback demands less resources ?

 

[Mnyb]

Embedded hardware support is great for flac , most streamers and whatnots can do it ?

 

[ZolaIII]

30 to 50% bandwidth saving (CD to higher quality) for offering perceptual same quality (hybrid lossy high quality WavPack vs flac) is a lot especially if you are a provider.

 

[ThatM1key]

There is a difference between the OGG media container and Vorbis the audio codec.

I know there is.

And even more compression is not needed , it would be single digit percent better , why with today's storage.

Flac can do multichannel but not DSD , but DSD is useless anyway just convert to multichannel pcm and then pack as flac .

Here i dont know how work intensive is wavpack to uncompress ? Flac is made so that the compression is the most computer intensive part , uncompress for playback demands less resources ?

(2L) Hoff: Innocence DSD 64 Stereo:
DSF: 200MB
WV DSD: 88MB
FLAC 24/44.1: 41MB

(2L) Hoff: Innocence DSD 64 Multi:
DSF: 650MB
WV DSD: 262MB
FLAC 24/44.1: 110MB

I guess you could but is it lossy technically when you go from DSD to PCM. But then again most DSD is DXD

 

[AnalogSteph]

Originally because WavPack was a symmetric codec that is it was as hard to decode as to encode.

Where as FLAC has always been asymmetric that is to say it's very computationally easy to decode but hard to encode.

WavPack now has an asymmetric mode but I'm not sure how it compares.

I just did some some comparisons using the latest Foobar2000 encoder pack from 2022-02-02 (didn't spot the asymmetric option though, maybe I'd have to update Foobar 1.5.7 as well). CPU is an i7-11700 (x2 tau boost, PL1/2 65 / 224 W, actual turbo clock seen 4.7 GHz).
FLAC 1.3.4 -5: encode ~460x, decode 1375x
WavPack 5.4.0 normal: encode ~290x, decode 373x
WavPack 5.4.0 fast: encode ~335x, decode 504x
File sizes for 48:52 worth of 44.1 kHz, 24 bit stereo audio were pretty much identical for FLAC -5 vs. WavPack normal, 561 MB in this case. Source audio was all in one file, so just one encoding thread, same on the decoding side.

This seems to align with official published results (with my machine being about twice as fast as a C2Q T9600 for this particular workload), except that FLAC decoding is even faster. It seems hard to make the case for WavPack if FLAC is faster even on the encoding side, unless you absolutely need on of its niche advantages.

Incidentally computational requirements is why Apple created ALAC they needed something relatively easy to encode for a fair compression ratio because they were doing it in real-time for Airplay 1 on low end hardware. None of the existing formats met their requirements so they invented a new one.

Interesting. Clearly much different objectives than FLAC, where the rationale was only having to encode once on fast hardware and saving computing power decoding on what might be mobile devices.

 

[bennetng]

WavPack has been asymmetric for a long time, including version 4.x which did not support DSD.

Monkey's Audio for example, has far slower decoding speed than WavPack, but was pretty popular in the 2000s.

Actually, I've been using WavPack for a long time, only jumped into flac due to flacCL, encoding speed was stupidly fast even on a Radeon HD5750 purchased in 2010. The same CPU (i3-540) in the same PC at the time was much slower in comparison.

For the lossy segment, Opus decodes much slower than for example AAC too.

Anyway, people should not use a format's popularity to determine it is good or not, otherwise things like JPEG should have been dead for a long time.

A WavPack related thread that some may find interesting:

32-bit floating-point (WAV) files "in the wild"

32-bit floating-point (WAV) files "in the wild"

hydrogenaud.io

 

[ZolaIII]

@bennetng interesting enough as it's old thread from time when FP rounding back caused integer overflows (fixed in IEE 754 revision later) but it also nedded some fixing on hardware FPU side. It whose and I don't know if it's still a case on ARM that it use integer written implementation (as originally it used it from the beginning). Unfortunately official public documentation is outdated and bad.
Anyway neither implementation nor computing requirement is a problem.

 

[mansr]

Incidentally computational requirements is why Apple created ALAC they needed something relatively easy to encode for a fair compression ratio because they were doing it in real-time for Airplay 1 on low end hardware. None of the existing formats met their requirements so they invented a new one.

Incidentally, ALAC is nearly identical to FLAC. The bitstream format is a little different, as are some mostly arbitrary constants, but the decoding process is essentially the same for both. Since FLAC predates ALAC by several years, they could easily have used that, but then anyone would have been able to play the files without using Apple hardware/software.

 

[threni]

Embedded hardware support is great for flac , most streamers and whatnots can do it ?

Yep. It was in Android from the start. I have no use for huge single files + cue as one of my use cases is copying files between devices, deleting them when I've heard them etc. Making compilations etc. If I rip a CD to a single file + cue I have a script to split them into the individual files. Unlike some formats, FLAC can handle this gaplessly (well, unless you're using something lame like VLC).
And I remember APE lol! The Opera of the audio world - why pay for something that's not better than the free alternatives? Converting APE to FLAC - another script that's handy to have laying around!

 

[bennetng]

WavPack is updated to version 5.5.0.

One moment, please...

One interesting thing is the website's introductory paragraph was modified according to some recommendations from a FLAC developer, because FLAC is not always possible to do this:

In the default lossless mode WavPack acts just like a 7-Zip compressor for audio files, including the preservation of all the headers and metadata, so the restored files are identical to the original.

The origin of the recommendations:

WavPack 5.5.0 Release Candidate

WavPack 5.5.0 Release Candidate

hydrogenaud.io

 

[fldmrshl]

I'm surprised WavPack isn't more popular too. I encoded my whole library in 320 kbps + correction. I almost don't use correction part, though.

I clearly can hear the difference between originals and 320 kbps "Spotify" OGG (they use the fastest encoding, I guess) in a blind test. I clearly hear the difference between originals and MP3 and AAC 320 kbps when I know what to expect. And I honestly cannot hear the difference between 320 kbps lossy WavPack and source CDs/FLAC.
Plus to that lossy WavPack surprisingly is the only lossy codec that does no clipping on compressed audio. With other lossy codecs I had to lower gain to -1.5 dB before encoding and still had clipping sometimes.

The best of all, WavPack doesn't use psychoacoustic model (as far as I know) to mask anything, so the only error you get in lossy 320 kbps .wv is statistical error, which is lower than dynamic range of perception of 99.9% of population and totally random - almost like a hiss on audiotape that masks the quietest sounds, just below the ability of any living being to recognize. Of course instead of theoretical 96 dB of dynamic range for CD you get maybe 80-90 dB which is totally neglectable, especially for highly compressed (talking about hard-limiters/brickwall dynamic compression) modern music. To me it's crucial no one (even scientists from Fraunhofer, Germany) plays tricks with my ears, because I like sound engineering and like to hear what's intended. I'd better tolerate indeed the hiss of audiotape than mp3/aac/ogg/opus/etc.

And the last, lossy WavPack is the only codec you can use if you're going to use Bluetooth connection and want to avoid 2x transcoding. If you use WavPack you're sure psychoacoustic model used only once on your audio -- when signal transferred from your phone to your headset. And if you use Bluetooth AptX it would be just two ADPCM-like encodings in a row, which is the best option -- compared to 2x transcoding with psychoacoustic model it's practically lossless.

 

[voodooless]

I clearly hear the difference between originals and MP3 and AAC 320 kbps when I know what to expect.

Would you be willing to prove this:

Is your audio system really ready for lossless?

Do an ABX test in your browser, and see if you can tell the difference between lossless and Spotify-compression quality audio

abx.digitalfeed.net

 

[Palladium]

Like the title says.

Main advantages over FLAC:

• Able to store and losslessly compress DSD stereo & multichannel.
• Album can be 1 Single with built-in cuefile.
• Create lossy files with a correction file to reconstruct the whole file again.
• Can losslessly compress better

So, that's four points that nobody gives a hoot to.

 

[ThatM1key]

So, that's four points that nobody gives a hoot to.

Yeah if your a boomer that listens to SiriusXM

 

[Multicore]

Re standards proliferation: de facto standards are MP3 and AAC [plus AIFF, plus WAV on the production side. flac to a lesser extent.]

MP4 containers are common too.

When I master for YouTube I put 2-channel 48k FLAC in MP4.

 

[fldmrshl]

Would you be willing to prove this:

Is your audio system really ready for lossless?

Do an ABX test in your browser, and see if you can tell the difference between lossless and Spotify-compression quality audio

abx.digitalfeed.net

I have no idea how to share the results link (and I don't think there is a way). When I tried to share on Facebook, it failed. And I don't use Twitter. So screenshots.

I'm 76% correct. There is a 1% chance that I could get this result by chance.

I didn't like the test though. I couldn't get the difference several times, and closed the test in the middle, feeling like whoever created this was trying to trick me. I think the choice of music in this test is awful -- I mean, almost all the compositions don't have a clear open hi-hat, there is a general lack of frequencies higher than 15kHz in all the tracks, and it looks like this choice of music was designed to be deceptive on purpose -- just to tell you "oh, you don't hear it, so you don't need it".

On a lot of tracks with more aggressive, saturated, rich music (metal, rock, post-hardcore) you'd hear the clear difference between "Spotify" OGG 320 and FLAC/CD. In real life, I think Spotify uses really bad fast encoding on their "High Quality" (OGG 320). AAC 320/256 and MP3 320 again -- in my opinion are more complicated. With MP3, I feel like I got used to it as a kid, and it's like a bad habit now. MP3 320 sometimes even seems more "natural" to me than FLAC/CDs because of it's aggressive saturated highs -- those FFT bands are like tiny gates/limiters, they make music sometimes "crispy" at high frequencies. But on tracks I like and love (mostly intense music: metal, hardcore, rock) I definitely hear the difference between lossy (I mean "psychoacoustic lossy") 320 and FLAC. And with lossy WavPack 320 (which is ADPCM-like coding, not psychoacoustic) I can't hear the difference with FLAC/CD at all, which is why I've encoded my entire library with it and use it without correction files.

Still, I have AirPods and my car has Bluetooth, and most of the time I listen to music in AAC 256 anyway, sadly. Convenience is more important in everyday life, which honestly makes me care less and less over time.


upd. I just noticed, result of the test says 256 AAC. That wasn't OGG 320 used on Spotify in this test as I was thinking initially. AAC 256/320 is the most preferable lossy codec of mine after lossy 320 WavPack. Generally in real live I don't hear the difference between AAC 256/320 and FLAC.

upd2. Also I wrote "I clearly hear the difference between originals and MP3 and AAC 320 kbps when I know what to expect" -- and you cited that. That meant no blind tests. That meant I listen to music I like and I love. I know how highs should sound like in those tracks (because I got FLACs). And highs (and maybe attack on drums or guitar) are honestly the only way you can practically make a difference between 320-256 lossy and lossless. Blind tests... I mean, I got 76% correct, and it was no OGG, it was AAC and music I don't like, and never listened to in FLAC/CD, and it was quite tedious lol

 

[voodooless]

Thanks @fldmrshl for taking the time to go through this ordeal! From the results, you can most likely hear a difference, so good for you and quite impressive

Yes, the test was AAC, not OGG, but close enough

 

[Palladium]

Yeah if your a boomer that listens to SiriusXM

Inaccurately insulting me isn't gonna address why your favorite format is DOA at adoption because its only solving nerd problems for nerds.

 

[ThatM1key]

Inaccurately insulting me isn't gonna address why your favorite format is DOA at adoption because its only solving nerd problems for nerds.

Calling all the points "Nerd Problems" is true but that's what all formats address just nerd problems. Innovation by "Nerd Problems" is how we get great stuff in life. Once nerd problems became mainstream.

I can argue why calibrate your system when you can just buy better speakers. Why compress WAV files when storage is cheap these days. Why stream files from your PC when you can just use a portable drive.

 

[Palladium]

Calling all the points "Nerd Problems" is true but that's what all formats address just nerd problems. Innovation by "Nerd Problems" is how we get great stuff in life. Once nerd problems became mainstream.

I can argue why calibrate your system when you can just buy better speakers. Why compress WAV files when storage is cheap these days. Why stream files from your PC when you can just use a portable drive.

1. Room interactions are infinitely variable
2. Compression has been cheap on CPUs, FLAC also has better metadata than WAV
3. Purely a UX issue than technical

Geez, you are getting inane.