TAK FAQ, The questions about the new codec — And it rhymes! Options

[Shade[ST]]

Q : Is this real?
A : It is. Try it for yourself; compress and decompress files, bit-to-bit compare, and all that other fantabulistic stuff. Look here for a download.

Q : Why should I use TAK?
A : If you're excited about it. The general concensus is that TAK compresses better and faster than FLAC, in its turbo mode. Albeit not being very supported right now, TURBO and FAST, (in some cases, perhaps, NORMAL) are candidates for hardware playback. How fast is it, you ask? Very fast. (thanks, Synthetic Soul)

Q : Will TAK be open-source?
A : Yes, TAK will be open-source, as soon as the code is ported to C or C++ and documented.

Q : When will TAK be open-source?
A : When it's done. More pressing issues for Thomas Becker are to create a binary containing basic compression and decompression, but that can be used as a dynamic library (i.e. for foobar2000 playback). He is currently working on a winamp plugin, and states that player support is the most important priority to him, for now.

Q : When its source is opened, what can we expect?
A : Discussed previously (a few months ago) was the license that Thomas would want to choose for his codec. He has not yet (or had) read the licenses for GPL, LGPL, et al., but seemed open to do so once his code was properly formatted and documented in C or C++ format. Also, he seemed to wish to publish a paper about his codec before opening the source, as to prevent plagiarism, and to give him proper attribution.

Q : Can I help the source release be made faster? I know delphi and C++.
A : Basically? No. Thomas wants to be the only one working on his source for the time being (cf : this post)

Q : So, what can I compress with it, then?
A : TAK 1.0 can compress any integer-format (up to 24bits per channel) PCM Windows Waveform file (.wav)

Q : How does it work?
A : TAK uses a forward prediction algorithm, the same type of algorithm that FLAC uses.

This post has been edited by Shade[ST]: Jan 29 2007, 22:25

[TBeck]

Thank you!

But i have some corrections:

' date='Jan 29 2007, 17:28' post='467944']
Q : Why should I use TAK?
A : If you're excited about it. The general concensus is that TAK compresses better and faster than FLAC, in its turbo mode. Albeit not being very supported right now, it is most likely to be very easy to decode on hardware, even in its most compressing modes. How fast is it, you ask? Very fast. (thanks, Synthetic Soul)

"even in its most compressing modes" is questionable. TURBO and FAST, maybe NORMAL are candidates for hardware playback.

' date='Jan 29 2007, 17:28' post='467944']
Q : When will TAK be open-source?
A : When it's done. More pressing issues for Thomas Beck are to create a binary containing basic compression and decompression, but that can be used as a dynamic library (i.e. for foobar2000 playback). He is currently working on a winamp plugin, and states that player support is the most important priority to him, for now.

My name is Thomas Becker, not Thomas Beck.

' date='Jan 29 2007, 17:28' post='467944']
Q : So, what can I compress with it, then?
A : TAK 1.0 can compress any integer-format (up to 24bits per channel) PCM Windows Waveform file (.wav) that does not contain additional metadata at the end of the file. That basically means any .wav file.

TAK can compress waves with additional metadata at the end of the file.

' date='Jan 29 2007, 17:28' post='467944']
Q : How does it work?
A : TAK uses a continuous-adaptation forward prediction algorithm. I'll get back to you to tell you what this means. Or someone can edit my post.

in "continuous-adaptation forward" continuous is wrong. The adaption is beeing performed (small)block wise, but not continuously.

Possibly it would be better, if you sent me the next revision via email before posting it.

Thank you

Thomas

[jcoalson]

' date='Jan 29 2007, 11:28' post='467944']it is most likely to be very easy to decode on hardware, for most modes.

I think this remains to be seen. I am very curious about this. the embedded cpus that are commonly used in music devices are totally unlike x86. they are usually not simd and if they are you have to write custom asm to use it. so the playback tends to be constrained by MAC performance in the inverse filter. but even TAK -turbo can use a higher order filter than FLAC -8.

Josh

[TBeck]

' date='Jan 29 2007, 11:28' post='467944']it is most likely to be very easy to decode on hardware, for most modes.

I think this remains to be seen. I am very curious about this. the embedded cpus that are commonly used in music devices are totally unlike x86. they are usually not simd and if they are you have to write custom asm to use it. so the playback tends to be constrained by MAC performance in the inverse filter. but even TAK -turbo can use a higher order filter than FLAC -8.

Because you have far more experience than i with hardware players, i will not contradict. My prefered formulation is "TURBO and FAST, maybe NORMAL are candidates for hardware playback" (see above).

But i assume, that at least non portable hardware players have more powerful CPU's. Maybe i am wrong?

BTW: I intend to add a new real TURBO codec to TAK, which will be significantly faster and only slightly less efficient than the current one. The current codec has not been built for maximum speed, but now has to serve as an all in one solution for fastest and strongest modes.

This post has been edited by TBeck: Jan 29 2007, 20:13

[rjamorim]

This thread really belongs at the Wiki :-P

[pepoluan]

I'm not too well-versed about TAK to write a HAWiki page about it. Best is of course Thomas. Maybe a collab of Thomas, Shade[ST], and Synthetic Soul (IIRC he rewrote TAK's readme).

[jcoalson]

But i assume, that at least non portable hardware players have more powerful CPU's. Maybe i am wrong?

some do, but not much more, and they have to do more at the same time, like run a network stack.