Do you think there will ever be a decoder that converts a WMA(Original Cd file) file to same size as mp3 but is lossless?

I suppose you can never say never, but I wouldn't hold my breath for this.

I think it'll matter less and less - storage is gonna keep getting cheaper, and at a certain point you won't really care if it compresses 75% or just 50%.

lossless is definately the future, the only question is when. In 5 years, when there's 250gb iPod, or more, it just won't matter all that much anymore.

Wma is not an original cd file, and it will never be possible to convert from wma to mp3 without loss.

He probably meant .wav

Then he is asking if there are ever going to be lossless codecs at "mp3 like" ratios, guess he means 4:1~10:1 lossless compression.

Should we simply say no? Current state of the art often doesn't even reach 2:1...


On the other hand, even if we have storage measured in tera*cough*tebibyte units, its always tempting to have 5 or 7 times more audio compressed lossy but "transparent" with the likes of mpc

Or do you think quantum computing could have an impact in audio compression? I know it will for encryption...

Some data (audio or not) can't be compressed at all.
This might be extremely (!!!) rare for a natural audio signal but not absolutely impossible.

If you could compress any data by just one byte, you could compress it again and again until you reach a zero byte file, which still deflates to the original data. That would solve some problems...

You're completely correct for random data, however, I cannot imagine what an audio signal that had absolutely no redundancy in it would sound like. It'd be a hideous cacophony, I'd imagine.

To the original topic: The fact that special purpose compressors can get a rough average of about 2:1 lossless compression on audio is pretty darned good, actually. I don't think there's any real chance of hitting, say, 8:1 average lossless compression on real-life audio anytime soon, much less 10:1 as MP3 is (very roughly).

Compression can be increased only so much. There is always a point at which all the redundancy is removed from the signal, and there's no further room to compress it, with any algorithim. Given that they know the type of data in advance, special purpose algorithms can achieve higher compression than general purpose ones designed to compress most anything, but there's always a limit.

And as cheerow points out, no algorithm can compress everything. It's mathematically impossible. Any given set of N bits can only represent 2^N possible sets of bits, and that's just the way it is.

I remember someone saying here that the advanced (yet to be seen) version of lpac, which had been choosen as the lossless mpeg4 audio codec, offers (little/much?) better compression ratios than what we have so far.
AFAIK nothing has been released yet...

edit: minor

I think the need for lossless compression nearing the level of lossy compression will simply never be strong enough to drive the research, even if it were technically possible to have the range of 10:1 lossless compression (which I doubt it is).

First, hard drive prices per GB are dropping (I've seen as low as $0.65 per GB), which means that the average music collection that most people have could easily be archived in current lossless compression with affordable hard drive space.

Second, with the continued improvement of lossy codecs, demand for lossless will drop somewhat, though this will also be countered by the drop in storage capacity prices. For now, most people can find 98%+ perceptual transparency somewhere between 128kbps and 200kbps, so other than special needs (repeated transcoding or bit perfect archival), this much more efficient encoding approach will transcend the need for substantially higher lossless compression ratios.

With storage becoming less of a problem I wonder if lossless will ever be interesting at all. Seems more logicla to choose a lossy one near transparency or simply stick to wav in my opinion...

I agree with scorlibran...i fail to see the motivation for doing heavy research into lossless compression.

Cheers!

Mazter

It would sound like white noise



Actually this could be rephrased to: compression ratios of statistical methods are limited to the entropy of the signal.

In a technique that isn't based on statistical methods such as Huffman tables etc, compression ratios aren't theoretically limited to anything. In addition to that, current compressors analyze only a microscopical fragment of the signal at a time and thus fail to see the large-scale redundancy present in the signal.

Imagine listening to a pop song where a computer-generated bassdrum hits 332 times and computer-generated hi-hat hits 1074 times. Plenty of redundancy, IMO, just like in the old MOD/FastTracker modules. Too bad this kind of redundancy is virtually impossible to utilize, but it actually might be plausible in the future.

So, to summarize, I believe that there is indeed some room for development in lossless audio compression algorithms.

If the so called cocktail-party-problem could be solved this would be a huge leap forward not only for audio compression (lossy/lossless) but for data processing in general. Yet this is just a dream, but hopefully one day... I think it is not absolutely impossible. We will see.

Lossless is tempting but far from practical even on a large hard drive. The fact that one needs to dedicate entire drive(s) for music and lets not forget backup - for each drive we need a secondary backup drive. All we are doing is adding drives - this is especially true for those who rip several albums a month. You could archive to DVD's but that is even more time consuming that imaging to another drive.

I have a 14GB collection on a secondary 120GB drive, most of it in MPC Q5. In lossless world this collection would probably reach around 100GB. There goes my secondary drive and I'll have to add a 3rd drive! (got better things to do). I backup the collection to the primary drive and it takes me around 20mins. Even if space was no issue I still think lossless is inefficient and that is not good for any technology.

IMO lossless is best used as an offline archival / transcoding solution.

I don't understand your last point. If space was not an issue, it seems illogical to care about the efficiency of a lossless codec.