Skip to main content
Topic: Lossless Compression Test (Read 54047 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Lossless Compression Test

Reply #25
I just convert some random "piano" album. length 33 min

Tak -5 = 94.2 Mb
Ape -High = 97.3 Mb

another one 72 min

Tak  262 Mb
Ape  266 Mb

I have tried a few soundtrack album and result is just like the above.

It depend, If I remember it will be around 2-7 Mb per album depend on how long and music type (quite one, I mean). Or simply, how "Clean" the signal is.

With noisier music and some music that have constant noise (like some jazz)  MAC usally have better compression. but different usually just about 1-2 Mb per album. 2Mb is pretty rare actually.

Random pick for rock = White zombie Astro-creep: 2000 length 52 min

Tak -5 = 342 Mb (358 854 414 bytes)
Ape -High = 341 Mb (357 577 462 bytes)

Slipknot : subliminal verses (This one is loud as hell) length 60 min

Tak = 425 Mb
Ape = 423 Mb

But maybe I have a lot of somewhat quite music and I remember that when I convert my library from APE to TAK it usually get smaller most of the time. So I'm a bit surprise when I saw your result. That's all 

Lossless Compression Test

Reply #26
Mac Insane mode added to the graphs

@buktore
thats the perfect answer, thanks
Who are you and how did you get in here ?
I'm a locksmith, I'm a locksmith.

Lossless Compression Test

Reply #27
Different genres of music will produce different results.

For TAK this means:

The files don't benefit from the high predictor orders of the presets p3 to p5. The advantage of p5 over p2 is only about 0.5 percent.

It's different for Josh Coalson's FLAC comparison:

Here the advantage of p5 over p2 is about 1 percent.

Could this mean you are considering music genre presets (rock, metal, jazz, classical) in future versions of TAK for encoding albums?

I have thought about it but am not sure if it is a good idea. While i could create specific settings for specific genres, which will work quite well on average, there will always be users with a file selection were it doesn't fit. Possibly it would be better to recommend genre specific settings and advise the users to try them on their file sets and to tell them how to tune them.

It depend, If I remember it will be around 2-7 Mb per album depend on how long and music type (quite one, I mean). Or simply, how "Clean" the signal is.
...
With noisier music and some music that have constant noise (like some jazz)  MAC usally have better compression. but different usually just about 1-2 Mb per album. 2Mb is pretty rare actually.
...
But maybe I have a lot of somewhat quite music and I remember that when I convert my library from APE to TAK it usually get smaller most of the time. So I'm a bit surprise when I saw your result. That's all 

This exactly reflects my own observations about TAK's efficiency in comparison with Monkey's audio. BTW: My own music collection behaves similar to yours.

Lossless Compression Test

Reply #28
A Man Eating Duck, could I egg you on to compressing your corpus with some OptimFROG and La settings?
I'm aware that it's a behemoth of a task, of course.  Maybe that's the reason why Synthetic Soul is remarkably discreet about it

Lossless Compression Test

Reply #29
A Man Eating Duck, could I egg you on to compressing your corpus with some OptimFROG and La settings?
I'm aware that it's a behemoth of a task, of course.  Maybe that's the reason why Synthetic Soul is remarkably discreet about it


Back when i'd done a few tests of my one (very limited, not even 1/10 of the size or scope of what Synthetic Soul was doing), I found that testing OptimFrog/LA would be a serious chore because of the encoding times.

However, testing OptimFrog at default (which is reasonably fast) should give us a good indication of where the "strongest" codecs are at.

Lossless Compression Test

Reply #30
I want to write here something to give some food for thought:

Because the space for Loslsess archiving/storing music is nowadays available without problems and without much expenses (big HDs, DVD+-R), those few tenth of percent or few percent of space saving by some Lossless formats compared to flac, don#t count very much (imo?!).

For some reason FLAC is used by a big majority of persons world-wide, see eg. the HA poll, clearly above 60% for flac, rest distributed over the other Lossless formats.

The reasons for this outcome are surely the en- and maybe even more important, the decoding speeds of flac, which translates also to low cpu power needed, which leads to battery savings and longer playtime on portables.
When we are at portables, and industry support, we meet FLAC again.

Somehow i want to appeal to the Lossless crowd, to think about customers power.
If we want in sooner or later future more (cheap/priceworthy) standalone and portable devices, which play Lossless, we should maybe concentrate on 1 preferred and used Lossless format. And, that is somehow obviously FLAC with already good acceptance by commercial industry.
If the industry has an easy decision to take, which Loslsess format to implement, they will be easier to convince, to just do it.
If the Lossless format usage is way more distributed, let's say: flac, wavpack, apple Lossless, ape, tak all used in similar percentages by end-users, industry will have difficulties to decide themselves, to implement any of Lossless. Or they take the "wrong" format, maybe 'Orange'-Lossless, which we wouldn't really want ?!
So, let's make FLAC the most used format, as it is, so we will get more and even more hardware support. portables, stand-alone HiFi, car-HiFi.

Lossless Compression Test

Reply #31
I like flac a lot but wavpack is more flexy, can run on modest hardware and has the BSD license which is better for business. The world needs a bridge between lossless and lossy and that is where the hybrid mode comes in.
In the end competition is good so bring on tak &co.. I have doubts about lossless becoming an encoding standard ala MP3 / AAC regardless of storage. It will stay as a transcoding platform and grow as an audiophile thing. Remember real world is not HA and the masses do not transcode they just want to play their songs and the more songs they can fit the better. To convince them otherwise, to adopt and default flac encoding into every common frontend and hardware device is a very difficult task - that is the only way to mass adopt lossless.
wavpack 4.8 -b3x6c

Lossless Compression Test

Reply #32
libFLAC is also BSD licensed.

if you go outside HA, the picture is different.  flac support is wide enough that I think distribution is driving now and lossless music being sold is almost exclusively in flac, with a handful of wmal places.  in legal trading it's flac and shn, and in usenet it's flac and ape.  apple may start selling alac, but they'd be better off being the first bug guy with flac since their lock-in is due to itunes/ipod ease of use, not codec choice.  unless something disruptive happens that seems to be where it's going.

Lossless Compression Test

Reply #33
Because the space for Loslsess archiving/storing music is nowadays available without problems and without much expenses (big HDs, DVD+-R), those few tenth of percent or few percent of space saving by some Lossless formats compared to flac, don#t count very much (imo?!).

Conversely, with processing power increasing, it could be considered that encoding (and decoding) time matters less and less. Also, you don't take into consideration that increased storage capabilities could be used for other types of content, such as high definition video: people used to archive 700 MiB XviD rips, now they're storing 4.7 GB or even 8.5 GB 720p/1080p rips.

I have divided encoding times by two on my €50 dual-core CPU (Ahtlon X2 4000+ EE). You could count on a ~400% increase in processing power with quad-core CPUs (by encoding 4 files at once). Now I can even encode audio CDs on my dual-core with wavpack -hhx6 (wavpack's highest setting) or even with Monkey's Audio (insane mode) faster than realtime.
An increased core count also changes usage patterns as well as the importance of other factors, such as decoding time. On a quad-core CPU, you could use two or three cores for encoding and not feel a thing, because you'd still have one or two more cores for whatever else you're doing. With that CPU, does it matter any longer if a highly-efficient codec has very slow decoding speeds, since you can now dedicate a whole core for that task? Would it matter that much if simple playback took 90% of processing power of a single core, since you'd have three more available? Mind you, I mentionned quad-core CPUs, but what about CPUs such as the UltraSPARC T2, which has 8 cores, each of which can run *8* threads concurrently? I know it's not a "desktop" product, but I don't think it's much of a stretch to say that we're headed for less GHz increases and more cores per CPU.

Take a typical situation: you just got back from the record store. You rip the CD, and while you're listening to the wav files, you're encoding them using the highest settings of a highly-efficient codec. By the time you're done listening to the album, your highly compressed files are ready for archival. From that point of view, is it so important to encode an entire album in 20 seconds instead of 20 minutes?

I'm all for improving encoding speeds, but that's mostly because it frustrates me to see that my CPU could run twice as fast if only applications were written with full support for all the relevant features (i.e. SIMD extensions) that make that CPU so great. It's so annoying to see that the lancer build of Ogg Vorbis runs roughly twice as fast as the stock encoder on my setup, yet no effort is being made to port it to GCC and thus make it run on anything other than Windows! Or take flake, for instance: on my setup, using only one core, it encodes Joni Mitchell's "Blue" in 15 seconds vs. 66 seconds for flac --best, yielding the same filesize, yet it is still being dismissed as a somewhat experimental encoder, not suitable for general use.

We keep disregarding newer, better software for reasons that are beyond my comprehension. Hardware is improving: let's make use of improved software with it.

Lossless Compression Test

Reply #34
@skamp: you miss the point, which is having a universal simple, but rock stable cross-platform format, with still excellent performance, at low cpu power, ie.: commercial stand-alone devices, HiFi-hardware made by companies, to buy at your local walmart, play your Lossless where you are at the moment, like you can do these days with mp3 in the lossy sector.
This means, in the Lossless area, squeezing space or speed out of Lossless, doesn't bring big progress. Of course, the nice thing in Lossless coding are the small differences between those different codecs, and used mathematics. Interesting from developers, scientists' points of views, but not for JoeAverage, exception: technical geeks, who are in higher concentration located at this music technilogy forum, of course. As User, I had to pin-point to the practical view, and that is working towards realization of Lossless decoding into company built hardware, portable or living-room HiFi.
Of course, competition between Loslsess formats is ok, as it might bring progress. Eg. flake and flac, but it shows also, many people want security from Lossless, not balancing on cutting edge. I am sure, the good things of flake or other codecs have been/are or will be implented to flac also, if there no other disadvantages.  Many people might use flac, because they can already play it on their Kenwood, Phatbox, in their Volvo car radio system, on Rockboxed devices. And so they trust in "their" flac developer, that he knows, how important the decoding/play factor is, ie. gaining industry support
You see, I myself used at least 3 Lossless formats each some time in past, as I was convinced, that format is the best Lossless compromise. (heh, every Lossless format is a compromise, if you count in every technical detail, but if you set preferences, you will get your personal favourite winner format for sure). I started with ape, then flac, then wavpack, then flac again, and I will stay with flac, I think (so far  )

Lossless Compression Test

Reply #35
I hear you on the matter of ubiquity, but you're assuming that the sole availability of a given format with relatively unrestrictive licensing terms and good performance is enough to bring the industry to support that format. In real life, the major players in that industry gang up together to create proprietary "standards" that they can license for a fee and make big bucks with (cf. the format wars).

FLAC has been around for years, yet its adoption among hardware manufacturers, while honorable, is still somewhat anecdotal. Look at Apple: they didn't have any legacy compatibility issues to take into account when they decided to bring lossless audio to their products (among which, the oh-so-popular iPod). They had the luxury of being able to choose whatever tech they saw fit. Did they choose FLAC? Noooo. They engineered their own inferior codec.

The only support you can realistically hope for, is that coming from vendors who are starting to bet on cheap, open technology, that they can sell to masses and thus still make good money from. Asus, with its EeePC, comes to mind (then again, why would you use lossless codecs on slow hardware with very little storage?).

FLAC has benefited from pretty good adoption among an independant public (such as independent artists and online shops, but mostly illegal BitTorrent trackers); I'm not sure why. Maybe because it was among the first codecs, and had the most features early on?

Anyway, I question the importance of wide commercial support for any lossless codec, simply because of the very purpose of lossless encoding (bit-for-bit archival). It doesn't really matter if your iPod supports your lossless codec of choice, because lossy codecs are much more appropriate for that target (relatively fast decoding, very small filesizes, while providing transparent playback). It doesn't really matter if your standalone DVD player supports FLAC, because the industry will keep pushing their own technology onto you (DVD-Audio, SACD, Blu-ray, etc...). And it doesn't really matter if your online shop sells FLAC or WavPack files, either, because converting them into your favorite format is easy and fast enough.

Lossless Compression Test

Reply #36
mp3 won the lossy war because consumers chose it, flac is the consumer choice for lossless so when drive size continues to rise there will be much more interest in it, i know people that say defensively "there's no way you can hear a difference between LAME mp3 and flac" but as drive size rise, they'll be answering themselves "well i have the drive room, so why not?"

Lossless Compression Test

Reply #37
I'm not so sure CoyoteSmith. Mp3 won the war because there was no real competition. Now there is competition, most consumers aren't interested. Quite right too. Thanks to Lame (no thanks to Franhofer), mp3 is still holding its own competing with the latest codecs.

There are some people on this forum who have abx'd enough samples to not trust a lossy codec, but they are a tiny minority. Consumers don't abx. Consumers don't know or care about lossless. I only know a couple of people who have even heard of lossless audio codecs. I don't know anyone who knows what replaygain is. Personally, I'd much prefer to see replaygain/gapless hardware support for mp3 than lossless support. Thats just me though.

Edit: I use wavpack for my archival requirements. Viewing the graph from the OP I'm still happy with my choice.

Lossless Compression Test

Reply #38
Consumers don't abx. Consumers don't know or care about lossless.

Consumers don't know about transparency either. Most people are convinced that there has to be an audible difference between original and mp3.
So, if you can convince them (which isn't easy either) that lossless really really is lossless they might actually be more likely to use it than the HA crowd.

Lossless Compression Test

Reply #39
Gaekwad2: I disagree that most people think there will be a difference between lossy/lossless. They generally don't think about such matters. go out tomorrow and ask your neighbours which is their lossless codec of choice. watch their eyes glaze over before they shut the door.

Lossless Compression Test

Reply #40
go out tomorrow and ask your neighbours which is their lossless codec of choice. watch their eyes glaze over before they shut the door.

They'll probably remind their 10 year old daughter never to talk to strangers as well, especially the creep next door

Lossless Compression Test

Reply #41
go out tomorrow and ask your neighbours which is their lossless codec of choice. watch their eyes glaze over before they shut the door.

They'll probably remind their 10 year old daughter never to talk to strangers as well, especially the creep next door

well it just happens that with my line of work i do meet plenty of different DJ personas, mostly they don't have any clue about what transparency means (usually it's something to do with 192 kbps cbr mp3s), they don't know about replaygain, flac (and entire lossless compression) seems like sci-fi to them..., so how about expecting something like that from your neigbour? 
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung

Lossless Compression Test

Reply #42
I want to write here something to give some food for thought:

Because the space for Loslsess archiving/storing music is nowadays available without problems and without much expenses (big HDs, DVD+-R), those few tenth of percent or few percent of space saving by some Lossless formats compared to flac, don#t count very much (imo?!).

For some reason FLAC is used by a big majority of persons world-wide, see eg. the HA poll, clearly above 60% for flac, rest distributed over the other Lossless formats.

The reasons for this outcome are surely the en- and maybe even more important, the decoding speeds of flac, which translates also to low cpu power needed, which leads to battery savings and longer playtime on portables.
When we are at portables, and industry support, we meet FLAC again.

Somehow i want to appeal to the Lossless crowd, to think about customers power.
If we want in sooner or later future more (cheap/priceworthy) standalone and portable devices, which play Lossless, we should maybe concentrate on 1 preferred and used Lossless format. And, that is somehow obviously FLAC with already good acceptance by commercial industry.
If the industry has an easy decision to take, which Loslsess format to implement, they will be easier to convince, to just do it.
If the Lossless format usage is way more distributed, let's say: flac, wavpack, apple Lossless, ape, tak all used in similar percentages by end-users, industry will have difficulties to decide themselves, to implement any of Lossless. Or they take the "wrong" format, maybe 'Orange'-Lossless, which we wouldn't really want ?!
So, let's make FLAC the most used format, as it is, so we will get more and even more hardware support. portables, stand-alone HiFi, car-HiFi.


From what I've seen so far on the net I think that in the near future FLAC will become the standard for lossless audio in a similar way to MP3. Until recently, lossless file where only to be found on the user's collection of converted CDs, but now more and more files, concerto recordings, etc appear on the net, some portable devices start supporting them, ... and in each time it's FLAC.

Let's face it, it's probably not the most efficient encoder but the others are not much better, it's less unknown than the others, open source and free (a BSD license, so there are no royalties to pay).
MP3 started in the same way (i.e. free, in the beginning you had nothing to pay in order to develop your own software decoder, but after a few years and the growing success of the format, patent holders started to realize that they could make a lot of money and modified their policy).

The only restriction I see with FLAC, and most others lossless formats for that matter, is a tagging support far below what's available for .mp4 or what id3v2 tags provides for .mp3.

I must say thought that the current trend in CD converters adding separate .cue files gave me somewhat to think on the subject: why do these file formats need to support tags at all? If players can decode separate .cue files for titles and timing information, they could do the same for tags.

I dunno the exact capabilities of id3v2 or .mp4 containers but imagine that one can create independently from the audio files a file containing the album cover scans, title and lyrics for each track, etc. Codec developers could concentrate on the audio side while independently other developers concentrate on the tools (and source code libraries) for reading and editing such files or even downloading them from a freedb like central database.

André

 
SimplePortal 1.0.0 RC1 © 2008-2019