Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: TAK 1.0 - Final release of the new lossless codec (Read 252056 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

TAK 1.0 - Final release of the new lossless codec

Reply #125 –
Quote from: Lyx on
To make it quick and brutal: *currently*, we could still be happy with mp3 and shorten (exception: narrowband). All "modern" codecs do not offer a *significant* advantage regarding speed/compression.

This isn't 'brutal'. I am aware of the (limited) practical relevance of the compression improvements achieved since Shorten.

But regarding the speed, i have to contradict. For instance TAK is encoding much faster than FLAC. Ok, if you say FLAC -0 is enough, than speed differences are irrelevant.

Quote from: Lyx on
However, you dont arrive at a "significant advantage" just overnight. Sure, in some rare cases a technological revolution happens overnight, but thats not how it works most of the time. Thus, the work you put into your codec, as well as all other "modern" audio codecs, is a "longterm-investment" which will someday lead to a result which indeed offers a significant advantage over mp3/shorten. But for now, the masses dont benefit much from it.

I doubt, that we can expect a significant improvement as a cumulation of many small future improvements. I am convinced, that it would need some 'revolution' to get considerably futher.

But this is only based upon my intuition, i am not able to prove it...

Fortunately some people are already happy with the insignificant improvements achieved, otherwise there would be little motivation to release new codecs (I have been working for many years on TAK, without any publication in mind. Without the motivation resulting from the feedback of potential future users, i never would have started the work on a release.)

  Thomas

TAK 1.0 - Final release of the new lossless codec

Reply #126 –
Quote from: Synthetic Soul on
I think you only need to see the excitement that TAK has generated, or even the improvements in FLAC 1.1.3 over 1.1.2, to know that users are very keen to squeeze just a few more MB out of a file (if there is no speed penalty).


Yes, i'm aware about that. I wasn't questioning that there is an interest in it. I was just questioning the actual benefit from a rational point of view. I know that most people either simply do what they're told, or tend to become obsessive about single aspects of things while ignoring the rest (which possibly on the global scale legitimates the questional path which technological development is following currently). I'm aware that most people dont care about this or disagree with it - the opinions of others however are secondary to me. I'm primarily interested about objectivity - so, whats really the case, not necessarily what everyone else likes.
Quote
I doubt, that we can expect a significant improvement as a cumulation of many small future improvements. I am convinced, that it would need some 'revolution' to get considerably futher.

I'm not certain if i possibly worded myself a bit ambigious in my prev post, so i thought maybe i should make a bit more clear what i meant. I wasn't suggesting that FLAC or TAK or whatever will someday offer a significant compression gain (at reasonable speed). Technological projects aren t islands. They are built on previous knowledge. Others in turn will learn from those again, and so on. This partially is even the case for breakthroughs - contrary to popular belief, many breakthroughs dont just come out of nowhere out of pure chance. More often it is the case that the various puzzle-pieces are already there and it just needed someone to connect them or do the next step.

So, no matter if TAK will become and important player or not. For certain you will have learned something from it and did therefore progress. And possibly it will also contribute ideas and information to other developers which will then in turn do the next step. Thus, regardless of if TAK as an "implementation" suceeds or not, its ideas contribute to a longterm-gain. Or to put it simply: TAK is a small part of technological evolution.

- Lyx
I am arrogant and I can afford it because I deliver.

TAK 1.0 - Final release of the new lossless codec

Reply #127 –
Quote from: Lyx on
Quote

I doubt, that we can expect a significant improvement as a cumulation of many small future improvements. I am convinced, that it would need some 'revolution' to get considerably futher.

I'm not certain if i possibly worded myself a bit ambigious in my prev post, so i thought maybe i should make a bit more clear what i meant. I wasn't suggesting that FLAC or TAK or whatever will someday offer a significant compression gain (at reasonable speed). Technological projects aren t islands. They are built on previous knowledge. Others in turn will learn from those again, and so on. This partially is even the case for breakthroughs - contrary to popular belief, many breakthroughs dont just come out of nowhere out of pure chance. More often it is the case that the various puzzle-pieces are already there and it just needed someone to connect them or do the next step.

No contradiction from my side. Not regarding your description of technological evolution...

But there will always be limits, which can not be overcome. The question is, how close lossless audio compression has come to those limits. Nobody knows, but my intuition tells me: quite close. Therefore i possibly preferred the term 'revolution', because i would regard any considerable improvement as a miracle...

TAK 1.0 - Final release of the new lossless codec

Reply #128 –
Quote from: Lyx on
Yes, i'm aware about that. I wasn't questioning that there is an interest in it. I was just questioning the actual benefit from a rational point of view.


I think it's entirely rational to want to use to best technology available.  You can look at just about any technology (mature or otherwise) and find efforts to make it as good as it can be.  Lossless audio technology is no different.

Quote
So, no matter if TAK will become and important player or not. For certain you will have learned something from it and did therefore progress. And possibly it will also contribute ideas and information to other developers which will then in turn do the next step. Thus, regardless of if TAK as an "implementation" suceeds or not, its ideas contribute to a longterm-gain. Or to put it simply: TAK is a small part of technological evolution.

- Lyx


I find this to be pretty meaningless.  Whether TAK plays a small role or becomes a/the codec of choice is pure speculation.

TAK 1.0 - Final release of the new lossless codec

Reply #129 –
Quote from: Mr Bungle on
I find this to be pretty meaningless.  Whether TAK plays a small role or becomes a/the codec of choice is pure speculation.
Guess why he said "regardless whether..." ?
He wasn't implying that TAK was going to have a small influence.

Though this still is OT.

TAK 1.0 - Final release of the new lossless codec

Reply #130 –
Yeah he was...

'TAK is a small part of technological evolution'

is what he said.

OK that's enough from me on this off-topic.

TAK 1.0 - Final release of the new lossless codec

Reply #131 –
hello !
i am not using lossless compressors at all, but anyway i am impressed of your work man !
thats really something big !
i have also some questions to you
Quote from: TBeck on
- Piping support.


you mean, it can read from stdin and output to stdout ? i havent found any info in documentation about it.
if you didnt mean that, is it in plans ?

Quote
- MD5 audio checksums for verification and identification.


to make more use out of it, let user get the stored fingerprint by some command like getmd5 etc...

Quote
- Unicode support.


does it accept unicode filenames ?

/edit: fixed quote tags

TAK 1.0 - Final release of the new lossless codec

Reply #132 –
Your quotes are from the to-do list.  I think this may answer some of your questions.

Yes, piping is reading from STDIN or writing to STDOUT.
I'm on a horse.

TAK 1.0 - Final release of the new lossless codec

Reply #134 –
Current progress

Back with my progress reports!

What i am working on:

WinAmp playback plugin

Done.

SDK

The Software development kit will consist of a DLL and header files for Delphi Pascal and C. Most probably it will provide not only decoding but also encoding functionality. Hopefully other developers will use it to write plugins for audio applications not supported by myself.

Currently i am in contact with one experienced developer to discuss my interface. After the finalization of the interface and some testing performed by myself, i probably will sent the SDK to some interested developers for external testing.

Decoder improvements

Even more error tolerance for damaged files.

Encoder

The additional evaluation level EXTRA is back (last found in YALAC 0.11). Main reason: I can not make any single encoder option (of the options dialog) available to the developers, because those options will change in the future, if i improve the encoder. Those internal changes have to be hidden, otherwise i would have to update the encoder interface on every change. As a compensation i reintroduced the evaluation level EXTRA.

Then we have the 5 presets TURBO, FAST, NORMAL, HIGH and EXTRA which can be combined with evaluation levels STANDARD, EXTRA and MAX (Should be regarded as 'Insane'). This should be enough options.

Probably i will also remove access to the internal options from the TAK applications...

That's all for now.

  Thomas

TAK 1.0 - Final release of the new lossless codec

Reply #135 –
Quote from: TBeck on
Then we have the 5 presets TURBO, FAST, NORMAL, HIGH and EXTRA which can be combined with evaluation levels STANDARD, EXTRA and MAX (Should be regarded as 'Insane'). This should be enough options.

Probably i will also remove access to the internal options from the TAK applications...
I agree. The huge options I had with OptimFROG always give me a headache. Too many things to fine tune. Give users some (hoepfully) tuned presets and do not expose everything.

BUT

I still recommend you give something like "custom parameters" where expert users can just type in their parameter preferences, perhaps maybe to tune the compression for their particular case. You don't have to provide a radio-button interface. A textbox should suffice.

TAK 1.0 - Final release of the new lossless codec

Reply #136 –
Quote from: pepoluan on
Quote from: TBeck on
Then we have the 5 presets TURBO, FAST, NORMAL, HIGH and EXTRA which can be combined with evaluation levels STANDARD, EXTRA and MAX (Should be regarded as 'Insane'). This should be enough options.

Probably i will also remove access to the internal options from the TAK applications...
I agree. The huge options I had with OptimFROG always give me a headache. Too many things to fine tune.

Well, the user doesn't have to use all the options. But nevertheless you are right,

- because first impression is "Puh, so many options..." and
- because even less experienced (and less technically interested) users may be tempted to try the advanced options, just to make sure that they really are using the optimum settings...

Quote from: pepoluan on
Give users some (hoepfully) tuned presets and do not expose everything.

The current presets are the result of many months of evaluation, performed by me and the hard working early (YALAC-) testers. They should be very well tuned.

Quote from: pepoluan on
BUT

I still recommend you give something like "custom parameters" where expert users can just type in their parameter preferences, perhaps maybe to tune the compression for their particular case. You don't have to provide a radio-button interface. A textbox should suffice.

Yes, this would be nice to satisfy technically interested users. But i doubt, that this will be possible...

After i have done my homework (player support and SDK) i will continue the work, that is most fun for me: Optimization of speed and compression efficiency. I may tune the current all-in-one codec by adding new options and also add new codecs specialised for maximum speed or compression efficiency.

I already wrote about my new dedicated TURBO codec, which will hopefully decode significantly faster and also provide even better compression than the current TURBO preset...

All those optimizations will affect the individual encoder options: some will be removed, some added and some modified. If i wanted to expose all those options, every new optimized version would be likely to be incompatible (detailed options wise) to the previous one.

But technically interested users may join the "evaluation and optimization" club, when i ask for testers which help me to tune the new codecs or options.

TAK 1.0 - Final release of the new lossless codec

Reply #137 –
I'm getting:

C:\Encoders>takc puede.wav
Command line error: invalid mode

C:\Encoders>takc puede
Command line error: invalid mode

C:\Encoders>takc -mode -e -pN puede.wav
ommand line error: invalid mode

Any ideas?

TAK 1.0 - Final release of the new lossless codec

Reply #138 –
Quote from: kwanbis on
I'm getting:

C:\Encoders>takc puede.wav
Command line error: invalid mode

C:\Encoders>takc puede
Command line error: invalid mode

C:\Encoders>takc -mode -e -pN puede.wav
ommand line error: invalid mode

Any ideas?

This is the right syntax:
Code: [Select]
takc -e -pN puede.wav

Hope it works.

TAK 1.0 - Final release of the new lossless codec

Reply #140 –
Quote from: TBeck on
But there will always be limits, which can not be overcome. The question is, how close lossless audio compression has come to those limits. Nobody knows, but my intuition tells me: quite close. Therefore i possibly preferred the term 'revolution', because i would regard any considerable improvement as a miracle...


I disagree somewhat with this statement.  There have been many times in technological development that people have said "its not possible."  Even though it may be a ways off, we will all see lossless files the size of lossy files.

TAK 1.0 - Final release of the new lossless codec

Reply #141 –
I knew of TAK in passing, but didn't really pay much attention to it, in large part because I've been cutting back on my foruming a lot over the past year (takes up too much of my free time). I came across this post yesterday, read all the old posts regarding it's development, and today did some testing of my own using foobar2000 v0.9.4.2 to transcode and tag. Let me say that I am incredibly impressed by the work you've done, and all for free no less. I've been using Monkey's Audio since 2003 or thereabouts, and just recently switched to FLAC for it's high level software and hardware support (have never been happy with it's compression level and speed at -8). What I would really love to see is TAK reach or exceed that same level of support some day because it is perfect for the role, giving the best of both worlds between APE and FLAC. High compression, fast decode, and depending on compression level fast encoding too. Hopefully playback support will come to foobar2000 soon. Thanks for all your hard work, perserverance, and a really great codec! Take care!

TAK 1.0 - Final release of the new lossless codec

Reply #142 –
Quote from: spockep on
Quote from: TBeck on

But there will always be limits, which can not be overcome. The question is, how close lossless audio compression has come to those limits. Nobody knows, but my intuition tells me: quite close. Therefore i possibly preferred the term 'revolution', because i would regard any considerable improvement as a miracle...

I disagree somewhat with this statement.  There have been many times in technological development that people have said "its not possible."  Even though it may be a ways off, we will all see lossless files the size of lossy files.

Did i say "its not possible."? It would be a wonder for me, but wonders seem to happen now and then...

Quote from: Digisurfer on
Hopefully playback support will come to foobar2000 soon.

Probably you will not have to wait very long...

Quote from: Digisurfer on
Thanks for all your hard work, perserverance, and a really great codec! Take care!

Thank you very much!

  Thomas

TAK 1.0 - Final release of the new lossless codec

Reply #143 –
Quote from: TBeck on
Quote from: spockep on

Quote from: TBeck on

But there will always be limits, which can not be overcome. The question is, how close lossless audio compression has come to those limits. Nobody knows, but my intuition tells me: quite close. Therefore i possibly preferred the term 'revolution', because i would regard any considerable improvement as a miracle...

I disagree somewhat with this statement.  There have been many times in technological development that people have said "its not possible."  Even though it may be a ways off, we will all see lossless files the size of lossy files.

Did i say "its not possible."? It would be a wonder for me, but wonders seem to happen now and then...

Quote from: Digisurfer on
Hopefully playback support will come to foobar2000 soon.

Probably you will not have to wait very long...


foobar2000 + TAK will be the sweetest thing.

TAK 1.0 - Final release of the new lossless codec

Reply #144 –
Current progess

While waiting for feedback for my SDK, i spent some time on something different:

As i wrote earlier, the individual encoder options will be removed from the Compression options dialog of the public TAK releases (They will still be accessible in special evaluation versions for some selected testers).

Unfortunately the options dialog looked quite empty after this... But fortunately i've got a hopefully good idea to use the new space.

Now there is a comparison table with 15 rows, each containing data of one combination of preset and evaluation level:

Turbo + Standard
Turbo + Extra
Turbo + Max
...
Extra + Standard
Extra + Extra
Extra + Max

Each row contains 3 columns: Compression, Encoding speed and Decoding speed.

If you select a Preset+Evaluation combination (either by cklicking into the table or by using the preset and evaluation buttons), the corresponding row will be selected and it's data will be used as reference.

Example:

After selceting Normal + Standard you will see this:
Code: [Select]
                    Compression   Encoding speed  Decoding speed

Turbo  + Standard     - 1.26 %      * 2.46          * 1.14
...
Normal + Standard       0.00 %      * 1.00          * 1.00
...
Extra  + Max          + 0.75 %      * 0.12          * 0.75


Normal + Standard now is the reference. How to interpret the data:

- By using Turbo + Standard you would loose 1.26 percent compression. Selecting Extra + Max would increase the compression by 0.75 percent.
- Turbo + Standard would encode 2.46 times faster.
- Extra + Max woud achieve ony 0.75 of the decoding speed.

Important: This comparison is based upon the data from my primary test file set! Other files and systems (different cpus) can give quite different results.

But nevertheless i think it's a helpful feature, especially for new users. It can help to choose the optimal preset for your application.

What do you think?

TAK 1.0 - Final release of the new lossless codec

Reply #145 –
I think the idea is quite sound, but I would be concerned about putting specific figures to the table.

Is it possible that you could do something similar with a rating-type system?  maybe even coloured indicators (like a litmus-type scale)?

Another option:  ask for volunteers to provide data, so that you can at least say that the figures were calculated using a huge corpus involving a tremendous variety of files.

I dunno.  If you are happy that your current corpus is quite indicative of expected results (I guess you must have quite a good idea by now, comparing it to the alpha and beta tester's results) then I guess it would be fine.

I just wonder whether giving users specific figures may give you tech support headaches.

As I say though, I think a table is a nice touch.
I'm on a horse.
 

TAK 1.0 - Final release of the new lossless codec

Reply #146 –
Quote from: Synthetic Soul on
I think the idea is quite sound, but I would be concerned about putting specific figures to the table.

Is it possible that you could do something similar with a rating-type system?  maybe even coloured indicators (like a litmus-type scale)?

Hm, possibly a good idea. I will think about it...

Quote from: Synthetic Soul on
I dunno.  If you are happy that your current corpus is quite indicative of expected results (I guess you must have quite a good idea by now, comparing it to the alpha and beta tester's results) then I guess it would be fine.

My primary file corpus is quite representative but...

Quote from: Synthetic Soul on
I just wonder whether giving users specific figures may give you tech support headaches.

... very good point! This has also been my most important concern...

I am really not sure. I suppose, this table could be very useful for new users, but especially new users will be most irritated, if their results don't match the table data...

TAK 1.0 - Final release of the new lossless codec

Reply #147 –
I think the comparison table is a good idea. While accuracy is something of an issue, I think as long as you include a disclaimer saying where the figures are derived from then you'll be fine. It would certainly be helpful to many users who are trying to decide which preset to use.

TAK 1.0 - Final release of the new lossless codec

Reply #148 –
Quote from: TBeck on
Quote from: Synthetic Soul on

I think the idea is quite sound, but I would be concerned about putting specific figures to the table.

Is it possible that you could do something similar with a rating-type system?  maybe even coloured indicators (like a litmus-type scale)?

Hm, possibly a good idea. I will think about it...

Okay. Now no more absolute values:

Code: [Select]
                    Efficiency   Relative      Encoding speed  Decoding speed

Turbo  + Standard      0 %         - 63 %        * 2.46          * 1.14
...
Normal + Standard     63 %            0 %        * 1.00          * 1.00
...
Extra  + Max         100 %         + 37 %        * 0.12          * 0.75

'Efficiency' is the portion of the range defined as the difference of the weakest and the strongest preset. Not exact, but my english is too limited...

An example:

Weakest preset: 55.00 % (absolute)
Strongest preset: 53.00 % (absolute)
Range: 55.00 - 53.00 = 2.00 % (absolute)

If Normal + Standard achieves an absolute compression of 53.74 % then it's efficiency is:

  (55 - 53.74) / 2.00 * 100 = 63 %

'Relative' is similar but relative to the selected (reference) preset.

Possibly i should replace speed values less than 1 (slower) with the reciprocal: Instead of '* 0.75 write '/ 1.33'. This means 'is 1.33 times slower'.

Quote from: PabUK on
I think the comparison table is a good idea. While accuracy is something of an issue, I think as long as you include a disclaimer saying where the figures are derived from then you'll be fine. It would certainly be helpful to many users who are trying to decide which preset to use.

Fine. Unfortunately not many people are reading a disclaimer (me too...).

With the new approach, users have to read the documentation to know, what efficiency means... That's probably a bit annoying but more safe...

TAK 1.0 - Final release of the new lossless codec

Reply #149 –
If you want to avoid all possible misunderstandings a scale from 0 to 10 (or similar) might be better than percentages though. It may seem unlikely that anyone could mistake numbers from 0% to 100% for the actual compression strength, but...