I'm trying Wavpack, and I don't see the point of the Hybrid Mode. The whole appeal of using it is that you can generate a correction file if you ever want to have the original file back, however, this correction file added to the WV file comes out to be about the same size as the lossless version anyway, so it's not like you are gaining any advantages at all?

Hybrid mode could be used without correction files, as simple lossy encoder.
lossy + correction is interesting in specific cases (like: listening lossy files on computer, and burning correction files only on optical media). But for pure efficiency, lossy + correction are less interesting than simple lossless files.

I use wavpack hybrid without correction files for transcoding purposes. Files take far less space than lossless ones and the perceptive quality of the lossy transcoded files is identical to the ones obtained from the original source. It's known to be less destructive than psycho accoustic based lossy codecs.

I've heard about this purpose of Wavpack's hybrid mode several times now. Since you can't restore the original file without the lossy file, what's the point of only storing the correction files on backup media? Burning both away would be more reasonable in my opinion.

Though, it is CBR and slow (if you enable the HQ switches (-h and -x)).

Radetz

and the lossy files are huge with no hardware support - I dont get it either. FLAC and transcode!

Sparing some money

yet

The HQ switches do not increase much the quality, it's about the same quality as current bitrate +50. I use 450 kbps with no HQ switch at all, for transparency reasons (metal is quite bit consuming). But 350 without HQ flags should be enough for most music. As far as I can remember, this is the recommended preset.

My entire music collection is in Musepack format. I've been thinking of change to Wavpack lossy because I have understood it doesn't suffer from artifacts like traditional lossy codecs, but instead it adds noise (but not perceptive if well configured I think). Is that right?

Yes, but Musepack artifacts would be hardly perceptible either.

A more recommended use for WavPack lossy would be for posterior transcoding to another lossy format.

What's your current mpc setting?

Recommended by who? WavPack docs (so THE recommendation I would say) says "At 320 kbps the quality is difficult for even critical listeners to distinguish from the original, and at 384 kbps WavPack becomes essentially transparent.".

That you do not use any quality switch is your own business. bryant (do a search if you wish) especially said that you should EITHER use -h or -x to get the most quality out of a 384 kbps lossy encode.

For high quality 384 kbps WavPack encodes, the recommended (proposed in the doc) switches are -b384x
For the highest quality 384 kbps WavPack encodes, the recommended (non proposed in the doc) switches are -hb384x

Both will be slow as hell though.

Radetzky

This is exactly what I do. I transcode from my WavPack lossy files to mp3 or to vorbis for portable use and sharing.
Furthermore, WavPack transparency gives me a warm fuzzy feeling when I directly listen my music.

You would rate WavPack lossy a better source transcoder than MPC (even with was is said at the bottom of this page?)

Radetz

That site is a crazy source of bullshit. Who is its author?

The author claims that, based on Den's findings, wavpack is only a better transcoding source when you are transcoding to Atrac. That's a huge load of crap and makes no sense at all.

WavPack is known for being a better source because it applies much less processing to the stream than psychoacoustic codecs.

I find hybrid mode very useful. I have my CD collection ripped in wavpack hybrid, which is available via ftp. A couple of friends have access to it, and they download the lossy part of tracks they want for appreciation purposes. If they like what they hear, they can always download the correction files and get the perfect cd image (cue sheets are available too).

Ok I guess if someone could compare the file sizes of the following it would make sense: (MPC quality 5 + FLAC) vs. (Wavpack lossy with a reasonable quality level in which noise is not audible + difference file)

If the first combination takes less space then what's the point indeed?

Have you even tried to see if the first combination takes less space?

FLAC alone compresses worse than WavPack, add a MPC file to it and the summed bitrate will probably go past 1mbps.

Did I open a can of worm? If you go one directory higher on the webserver I linked to, you will find the user name. Sound's like the guy who manages the sticky in the MPC-General section.

Radetz

edit: fixed a typo

Ah, the author and owner of the R** guide and trading group? :B

I'd better not even comment about it here, it wouldn't be safe.

Why is the rage?
I was just asking which would be larger not making any claims. An earlier post implied this that's why. Is the Wavpack hybrid lossy file + the difference file = size of Wavpack lossless file or any larger?

Same thing. There are switches you can play to make the correction file bigger or smaller. You can guess that if you opt for a high quality, high bit rate lossy file, the correction file is going to be rather smallish.

But still... IMHO, if you are going to keep the correction files, go FLAC. You will also get hardware support. WavPack hybrid is for that (special?) case where you can aford to spend a lot of space (384 kbps is more than LAME api) but, still, not THAT much (~800 kbps lossless). I too believe it gives better quality than MPC (with -hb384x), but they are not the same beast...

I find it sad (sad? I'm getting weird...) that WavPack doesn't get more recognition for it's hybrid more. I think it is because there is not enough "market" for it yet.

Radetzky

You obviously never saw me enraged



Well, your phrase looked like you were claiming the first combination took less space, and asked only what was the point of using wavpack hybrid. That's how I interpreted it, at least.




1-2% larger usually. Hardly makes much of a difference compared to pure lossless.

This is the only good reason I've seen mentioned.

What kind of "market" do you have in mind?
Used the way emtee does, it could be useful for online music stores. But, what are the chances of something like that?

So, like Lunatique, I can't really see the point of Hybrid mode for "normal" usage. Sounds very much like a niche product, to me...

One point would be that you'd need only one decoder. I thinks thats all actually, aside from the fact wavpack does compress better than flac. A matter of convenience. Personally wavpack is my favorite format. I don't use the hybrid stuff since I don't use lossy codec anymore.

Yet there is only a minimal amount of empirical evidence to support the notion that transcoding from lossy Wavpack is perceptively better than transcoding from HQ Musepack.



I think you misunderstood, Roberto. What the author states is actually correct. He is saying that all codecs except wavpack produced artifacts audible to Den when transcoding to ATRAC. However, it would be wrong to extrapolate this finding to other destination formats besides ATRAC based on this evidence alone.

wavpack hybrid would be good if had hardware support ... as it is, it is of no use for me ... i can have a flac (archived) and an MP3 (at my player and/or PC), and have the best of both worlds

Wavpack has lossless mode and hybrid mode, with or without correction files. Hybrid mode is a feature for those who might see some use in it. For those who don't Wavpack is till a very good and featurerich lossless format. So what's really the issue, what's the problem of having one more feature that you can use or ignore? The point of this thread eludes me.....

But better than none evidence that transcoding from HQ musepack is better than transcoding from lossy WavPack.



Why? Is Atrac so different compared to other psychoacoustic codecs?

Eludes me too. I think some people have too much free time on their hands.

I have transcoded wavpack to mp3, vorbis, mpc , wma and aac and the results are usually flawless. I've have never been able to abx any transcode using the -x switch.

Nobody is claiming that transcoding from HQ musepack is better. Due to the technical properties of the formats, it probably isn't. However, Musepack has other advantages, like blazing decoding speed which may be quite useful when you want to transcode something quickly and a transparency bitrate at least 25% smaller than Wavpack. If there is no significant difference in quality, Musepack's advantages may tilt the scale as transcoding source for some people.


Probably not (although we can't know for sure, since ATRAC is closed source, AFAIK); but this is still just an extrapolation and should be regarded with caution until verfied by actual evidence.
It is up to the destination encoder to decide how to handle the incoming data, so if the source is transparent, then any artifacts of the transcoded file must have been introduced by the encoder, which wasn't able to handle the lossy source optimally. Therefore, there could indeed be a difference among different destination formats.

Have you tried doing it with the -f switch too? I'm asking it because the ~25% increase in decoding speed given by the -f switch (though it's still 3x slower than Musepack) seems attractive to me for transcoding.

I don't use -f , though i've tried it and nothing bad happened. Measured quality does decrease with -f , though you can make up for it somewhat by using -fx.

The wavpack 4.2 decoder is on par with flac using -f and still very quick without it.

Using a PIII 550, Foobar with wv4.2 plugin uses 0-1% cpu using -bx, 1-3% using -h or -hx

A niche market indeed. I guess 80+% users are happy with MP3s WMAs @ 128kbps. Other more discernible users will want lossless. Few will have heard about relatively unknown alternatives like MPC and hybrid encoders.

I wouldn't say WavPack lossy is useless. 384 kbps with practically no artifact (and if you get one, it WILL be noise/hiss) or ~800 kbps for lossless vs. ~250 kbps for HQ lossy psychoacoustic codecs... WavPack lossy definitively has a place. It is just sadly not well known.

Radetz

I'm starting to think that the niche market Wavpack hybrid was intended for may be exactly what I'm looking for my needs........

I need to be able to rip my CDs and have them sound quite decent (I have dificulty ABXing LAME APS, so I'm not extremely picky) and be able to keep an archive, in case I need to burn the original into a file (DVD burner to the rescue). I've been using FLAC, but my harddrive is filling quick, and transcoding to multiple codecs just doesn't sound like much fun.... Does anyone have any reccomended tools or commandlines to use with EAC for this purpose?

So, does this mean that you haven't been able to ABX -f transcodes either?

I am aware of wavpack 4.2 speed enhancements, but I've been waiting for the final version to come out before I decide to entrust it with some archivals.

I never understood why FLAC having such limited hardware support seems like some kind of advantage. I don't use any of the listed hardwares that support it, and it seems none of the best selling hardwares support it either. Or, has FLAC's hardware support grown significantly since I last looked?

I started this thread not because I have too much time on my hand--in fact, I so busy that I barely have time to sleep (take a look at my website and you'll see why). I asked because it seems strange to offer the hybrid mode when most people who cared about sound quality would want to have that correction file, and the two files add up the the same size as the lossless version anyway. I do like the idea of having both files on a FTP server for friends to choose download without the correction file though.

when did you last look? http://flac.sourceforge.net/links.html#hardware

anyway, h/w support relative to mp3 is not good, but relative to other lossless codecs it's pretty good. aside from FLAC, you can play ALAC on iPod and that's about it.

Josh

I don't expect to able to abx transcodes with -f in normal situations, i'll try later anyway. Here is a summary of the 'measured' quality (avg noise) on 1 song:


-fb -60.92db
-fbx -63.73
-b -64.7
-bx -65.8
-hb -66.6
-hbx -66.97

These numbers (noise) don't mean anything, but the differences do and might = perceptual differences too in extreme cases. Basicaly the higher the number the better measured quality. You can see that -f degrades quality by around 4-6db compared with other modes, however -x puts in on-par more or less. You could pump the bitrate by 30-60k to compensate or use -fbx with 15-30k more bitrate and that will match quality. You can do these tests using the -n switch.

Also the output of wavpack 4.2b3 is bit identical to 4.1

Thank you for the tests.

Let me get something straight: The speed enhancements with 4.2 are all in the decoder or does the encoder have anything to do with them?

The encoder has nothing to do with them. You could encode in 4.1 and decode with 4.2, so you get the decoder speed boost. 4.2 has a marginal encoding speed boost for the non -x modes.

I just tried out Musepack, and I think it's really nice. It's commonly believed to be the best lossy encoding out there. But how does it compare to Wavpack's Hybrid mode in terms of quality (if you DON'T keep the correction file)?

The idea of lossless compression is that you are attempting to store something verbatim in the smallest possible space. If the hybrid mode produced two files that were smaller than a single lossless file, then obviously it would be possible to simply concatenate the two files and improve the lossless mode. Logically, the hybrid mode files must be bigger than a single lossless file; the idea is to have the extra overhead be as small as possible (like 1% or so).

I think of it the hybrid mode as lossless with additional flexibility. If I want to make some room on my HD I can delete the correction files of albums I don't care about anymore (or have backed up somewhere). When I copy stuff from DVD-R's to listen to I can copy just the lossy files, or I can copy the correction files too if I'm going to burn a CD. And the idea of sampling something online at 256 kbps, and then not having to have wasted that time to download the lossless version makes sense.

But, I believe the real benefit will be with portable use. Being able to have a 256-320 kbps version of music ready to copy directly from your lossless archive will always be faster than transcoding from pure lossless (although you could always do that also). The UPS man brought my new H120 today, and I am excited about working with the RockBox team on getting WavPack support in there, and I see no reason that Neuros support would be difficult either. And WavPack's low-complexity encoding will make it ideal for portable recording also.

BTW, version 4.2 is just a few days away. Thanks for all the support!

Drool....

Are there any current development on the Linux side too...?

If you mean creating say FLACs AND another format from EAC at the same time (which I think you're asking for), then MAREO and Flacattack can do this. However, transcoding from lossless to lossy is very easy - just use Foobar and do the transcoding ('convert' in Foobar terms) overnight (batch processing).

@bryant: Thanks for taking the time to give an elaborate answer about Hybrid mode, and not just defend it by saying that people that don't immediately see its usfulness, shouldn't bother. I'm not really familiar with WavPack, and was actually wondering what the point was, as no obvious areas of usage came to mind. But, now I see, and to me the "delete correction file if you're running out of space", is attractive, so I'll look more into it.


@rjamorim: How about we use the post count to measure "too much free time"? You could pretty much close down HA with that kind of attitude, as it implies that there's no need to discuss, learn and share opinions anymore.

As you've been along since the beginning, and has an impressive post count, you probably have alot of knowledge to share. But, judging by your attitude at times, it looks more like you're aiming for the title "The grumpy grandfather"...

To me, the elaborate answers by knowledgable people, is one of the things that makes this place attractive. It's also the reason why I read alot of different topics.

Thank you for the explanation Bryant.

I'm now trying to decide between Wavpack, Monkey Audio, and FLAC for my lossless compressions. They all seem to be really good, with bright futures in front of them. So hard to decide. . .. I think Wavpack isn't supported by dbpoweramp yet, and that's kind of important to me because I hate using foobar for sorting collections (I like graphical tab interfaces that's intuitive and neat). So, this leaves only Monkey Audio and FLAC. I've heard that FLAC is more forgiving when it comes to errors. Is that still the case now?

For lossy, I think I'll use MPC from now on, but the idea that you probably won't be able to hear the difference between MPC and a VBR encoded LAME at high quality, it makes me wonder if it's even worth it to give up the universal practicality of mp3.