TTA - very promising lossless WAV packer

[Bulat Ziganshin]

in 7z forum, someone given reference to True Audio Codec (TTA) http://www.true-audio.com/

i've compared this codec with wavpack, flac, lpac and sbc. it shows comparable compression ratio and good speed (to be exact, it has best time/space ratio among three open-source libraries in this testing). what is even more important for me, it has much smaller and simpler sources than any other lib that makes it easier to add its support to my archiver. so, again thanks to Johan, its next version will probably has state-of-art wav files compression. of course, this don't mean full support of MM compression - just one step in addition to using grzip MM mode

[Bulat Ziganshin]

moreover - version 2.0 of TTA, downloaded from http://www.compression.ru/download/sources/audio/t ta20.rar has a bit less compression for the same speed, BUT provides selection of compression modes - from -e1 to -e3. as you may already guess, the only mode remaining in modern TTA versions, is the FASTEST one. as a result, in -e3 mode it still remains a bit faster than wavpack/flac in their highest modes, but provides better compression. on those 2 wav files and small number of wav compressors i tested only lpac gives better results (but it is closed-source), while sbc, rar, uharc, wavpack and flac is worser than tta -e3

for those who don't have compilers i uploaded executable to http://www.haskell.org/bz/ttaenc.exe

[thometal]

Did you compare with "TAK"?

[Bulat Ziganshin]

no. but TAK is also closed-source. instead, i compared with Monkey, which shows better results. but its sources is about 5 tmes larger, which is important for adding its support to general-purpose compressor. otoh, on my 2 wavs monkey compressed 2% and 6% better than TTA

[Bulat Ziganshin]

according to Squeeze Chart, monkey seems to be the best available open-source compressor

[Stephan Busch]

Yes, that ape is really insane

What a pity that its author does not respond to my mails in order to be included in my hall of fame. Also no replies from Dmitry Shkarin and Markus Oberhumer. But however, it's their decision.

Cheers,
Stephan

[nimdamsk]

TTA is - no doubt - good. But not "state of the art".
TAK seems to be the most balanced audio compressor, but it's closed as it was mentioned.
Here is test sorted by encode time - http://www.synthetic-soul.co.uk/comparison/lossles s/index.asp?Sort=EncodeTime&Desc=0
As you can see TAK "turbo" and "fast" are both faster in compression rate and compression/decompression speed.
Here (http://softlab.technion.ac.il/project/optimizing_ tta_encoder/index.htm) you can find "boosted" multi-threaded versions of TTA, although they might be somewhat old.

[JohanDeBock]

OptimFROG is generally the best, but slow:
http://uclc.info

[JohanDeBock]

Here are some rather outdated timing results:
http://uclc.info/LossLess.pdf

[Bulat Ziganshin]

thanks, i've found your pdf and used it. can you please also include TTA 2.0 (see above fore sources and executable)? also, afaiu from wiki page about lossless audio codecs, your selection of music don't represent full range of various styles, so this comparision may be a bit biased?

OptimFROG isn't open-source so it's useless for my purposes but i plan to include ability to use external programs to compress data. it should be small revolution. now i use it to call ppmonstr, which makes a huge difference in text files compression

[Bulat Ziganshin]

btw, i've quickly analyzed monkey sources and found that compressing kernel something about 100 rb long. but managing it to the state of block-oriented compressor is still a serious problem, tta is much simpler in this respect

[JohanDeBock]

comparision may be a bit biased

All comparisons are biased.

[Bulat Ziganshin]

last news from the field

my own tests and various internet sources (say, http://studwww.ugent.be/~jdebock/LossLess.pdf) show the following ranking between lossless audio compression algorithms:

1. OptimFrog, LA, MPEG4 ALS. All are either closed-source or licensed to use only in MPEG4 products

2. Monkey, TAK. Monkey is symmetric, TAK comparing to Monkey has 3-4x slower compression but 3-4x faster decompression. TAK is written in Pascal, still closed-source, has large complex sources. Monkey is open-source, but not GPLed (each licensee should got permissions from author), source is also rather large

3. TTA 2.0
4. WavPack, SBC
5. TTA 3.3, FLAC

TTA/WavPack/FLAC are GPLed. Moreover, TTA 2.0 is faster than other GPL codecs and has very small sources (50kb, of which only 30 kb is codec itself)

so, for me the choice is obvious now i'm working on integration of TTA sources into my archiver. *current* version is available as http://www.haskell.org/bz/tta.rar , original sources at http://www.compression.ru/download/sources/audio/t ta20.rar