TAK 1.0.2 will come with a modified preset configuration.

The most important changes:

1) The fastest preset TURBO (-p0) is now using 8 instead of 16 predictors and compresses (on average) nevertheless better than FLAC -8 (with 12 predictors) and is about 6 times faster than FLAC -8 on my system.

The reduced cpu requirements should guarantee that this preset can be decoded on any hardware device capable to playback FLAC -8 (maybe even -5).

2) The default preset NORMAL (-p2) is now using 32 instead of 64 predictors. Compression is 0.3 percent worse on my primary sample set and encoding is 42 % faster.

I found the old default preset too slow compared with FLAC and WavPack. Now it's at least on par (speedwise) and achieves still considerably better compression ratios.

3) The strongest preset EXTRA (-p4) is now using 128 instead of 256 predictors.

I don't think that the rare files which can take advantage from more than 128 predictors can justify the heavily increased cpu requirements of higher predictor orders. But possibly 256 predictors will come back in form of an additional INSANE preset.

Here comes a table comparing compression and encoding speed of V1.0.2 and V1.0.1:

Test system: Celeron Tualatin (P3) 1200 Mhz
Program version: V1.0.2 eval 07
File set: rw (46 files))
Speed measurements without disk io.


And here a comparison of the compression of TAK Turbo and Normal with FLAC -5 and -8:


I hope you will like the new preset configuration.

The work on TAK 1.0.2 may take about 2 or 3 more weeks.

Thomas

Very nice.

Looks great. Waiting

"Since it is an asymmetric codec with very fast decoding it is an ideal candidate for hardware support, though."

the text is from the wikipedia entry.

can someone techincally explain this?

Presumably it means that it takes *significantly* less cpu time to decode TAK than it does to encode TAK.

In layman terms the code is optimized and therefore the decoding is very efficient and fast and requires less processing power. Therefore it's perfect for personal DAP's since they don't have much processing power. This also can translate into less battery drainage in said devices.

Oh and I am really looking forward to 1.02!! Thank you for the update. I have no doubt it will be up there with FLAC (in terms of users and hardware support) in the future.

seems like TAK is the lossless leading codec in a couple of months/years then. good, coz I am gonna transcode everything to it.

Can you elaborate? On what you base this assumption? FLAC has a very wide userbase and hardware support and TAK will have a hard time to match this if ever.
See the AAC versus MP3 case. AAC is technically better and pushed by Apple, but still cannot match MP3 regarding userbase and hardware support.

Well according to the polls ... FLAC is losing ground to WavPack sometime now because of speed and compression - once TAK supplies the other missing features I believe there will be a _MASS_ migration to TAK. And I really mean it... many people are loving TAK and eagerly following its progress. It's just a matter of time for it to catch up with FLAC and WavPack. There's only a couple of missing features in the do to list to catch up with the so called "popular" lossless codecs. But I still understand that NO lossless codec is that popular enough.

About the hardware, I consider FLAC to have limited to none hardware support. It really depends on what view you have. A U.S. citizen might have a certain "comercial" view of products, which are available and produced locally at affordable prices. Worldwidely speaking, it still VERY very limited on hardware. Since I don't have the US/European mentality that everything is at hand for a couple of dollars - this is what I consider. The FLAC hadware page has the most weird / spacy / boggling equipment I doubt would ever become as popular as DVD-Players have now. The "next big thing" for sure is the Blue Ray (BD-ROM) Player, so it's better that Sony include some lossless suport to it, because I won't be buying a weird 2-button device with a USB plug just because of FLAC.

What poll are you refering?
• 2006 HA lossless poll: FLAC/WV = 52% / 32%
• 2007 HA lossless poll: FLAC/WV = 59% / 22%




Since HA.org beginning, there was a mass migration from APE to WavPack/Flac, despite APE superiority in both compression ratio & encoding speed. So both features are apparently not the most important to a lot of people.


According to latest poll, 6 persons are apparently using it (OK, this argument isn't fair, as many people already voted before TAK's first release).



http://flac.sourceforge.net/links.html
Limited, indeed. None? nonsense...


The last big thing was SACD and DVD-Audio. They're both near-dead. I don't see how Blu-Ray would succeed for Audio (for video, it's a different matter). Anyway, no Blu-Ray Audio discs are planned as far as I know. It seems that manufacturers have understand that expensive multichannel HD-audio format has nearly no market except for classical music lovers maybe.



EDIT: the whole TAK future leading position should be removed and put somewhere else.

bourne you have a lot of faulty assumptions built in to that reasoning.

1. flac is losing ground... HA polls are not indicative of overall use, see http://www.hydrogenaudio.org/forums/index....st&p=492788

2. ...because of speed... flac is still faster than any other codec, encoding or decoding. anyway flac, tak, wavpack are all fast enough (in lower modes at least)

http://flac.sourceforge.net/comparison_all_cpudectime.html
http://flac.sourceforge.net/comparison_all_cpuenctime.html

3. what makes you think a few % compression is enough to enable mass migration? that's counterintuitive. wavpack has had about 2-3% compression advantage over flac for a long time.

4. ...no hardware support... what do you consider 'good' hardware support then and why do you think tak will exceed what flac has now? it's closed source and proprietary and as soon as it's open source and free, any of us can use whatever techniques we can glean from it to improve our own codecs. any techniques we cannot use due to patent issues will be a hindrance to tak adoption.

josh it's funny that in a post like this you prompted to reply it quickly to my assumptions - but the same self didn't bother to reply other posts in which I asked you why ffmpeg decodes FLAC at 64kbps.

About the poll. Although the numbers indicate that FLAC is still the leader, I am not measuring this by numbers. Did you read all polls comments? I get serveral "I was on FLAC but switched to WavPack", that's what I am counting. I presume there were much less people using WavPack and somehow it increased much more than expected. It is a way to say that FLAC lost ground yes. It is leader but then WavPack is not that too distant anymore. I am sure that once TAK issues are settled we will have another poll showing a lot of different numbers.

About speed: I don't do strict comparisons. You need to get the balance. I agree all these codecs are fast enough. Some are faster with certain settings, others are not. While some can be faster, they also can compress less or more. But then there is the seeking time, features, etc. TAKs proposal is to be like APE, with FLAC speed. Of course, making all ways to superiority. So it's not about "my codec encodes/decodes faster" - it's about a combination of things. I just wanna see FLAC compressing like APE.

About mass-migration: Certainly speed and compression isn't the main factor. But just take a look at the polls and see people commenting on how impressive speeds made up some people's mind. Once we have a combination of compression and speed people will buy this idea. But this is not all enough - I still didn't run into any FLAC gadget, as certainly as any TAK device. The day one just comes along my way I may change my mind.

About hardware support: MP3 has good hardware support. FLAC hasn't. Open source is not enough to get mass adoption. It's just a matter of which codec will fall in the grace of big companies to back it up. Could be FLAC, could be anything.

About Blue Ray players: Let's face it. It's the DVD-V replacement. Things will just keep going as they are - they're the ones who will be part of a home theater, sound system, etc. For the VIDEO part you better believe it. For the audio part, it will still play CD's with MP3, Audio CDs and legacy DVDs. SACD & DVD-Audio are dead projects because they're "audiophile" projects and this kind of thing companies realized it will never take over - why every DVD-Player has MP3 support? Because that is what is the hot thing now. I'm sure that when FLAC gets graced and introduced as supported in these players, many things will change. Audio in the future is much more file-management than PCM streams across media.

I don't know about your ffmpeg post, it's hard for me to keep track of everything. ffmpeg uses their own decoder impl (not libFLAC) which is simple but not optimized.

whatever flac is, I don't see any evidence in the rest of your post that points to a mass migration to tak. I know you're excited about tak, but making wild claims like this always drives threads off topic because they can't stand uncorrected.

Exuse my ignorance, but is there going to be some kind of DirectShow filter based off for TAK? Is it even possible?

josh, TAK looks very promising and I am just curious what the community will feel when its features get "completed". When I say mass-migration I mean that TAK will eventually chop some 20% off FLAC and WavPack in these polls. That's how I see it developing. Neither lossless codec has got hardware advantages on each other up to this point - those devices are not popular. So for me, and for _MANY_ people - still can't play FLAC/TAK/WV just like MP3 is played anywhere. You are interpreting that TAK will crush FLAC but this is not what I am saying. It has done very well in its early life and the future is exciting. What else can be improved within FLAC? The 2%+ compression ratio improved each year and no popular hardware support does not help. One credit to FLAC I give, it is that in the internet, it is very popular in P2P trackers.

Except of good compression rate and decoding speed any (lossless) codec must have good software support. I can play .flac files in almost any player under Linux. My preferred player is MPD (gapless playback and replaygain support most important for me) but it still has no support of Wavpack. I had some thinks about migration from FLAC to Wavpack (maybe even hybrid) but poor software support stopped me. So when TAK will play out of the box in Winamp etc. people will begin to think about migration. But not all, only some part. Even for p2p nets FLAC de facto standard in despite of non best filesize. It will be very hard to break synonym "lossless means FLAC" in mind of most people like now mp3 is just "shrinked equivalent of audio CD" for most.

Thanks TBeck for keeping lossless development busy... I like the new presets and "Compression is 0.3 percent worse on my primary sample set and encoding is 42 % faster" as default sounds very good. For lossless I'm using TAK now, because there are only advantages. Smaller files, better encoding speed (important here). No need for hardware support. No mobile hardware (currently it's only a car radio) needs lossless files here, I go for aac there. Looking forward to 1.02.

iGold, I don't know how about Linux, but there's a Winamp/tak input plugin.

I'm using TAK now because I wasn't hampered by legacy when I finally decided to switch to lossless. Sometimes, it's that simple.

It doesn't, it just doesn't display the bitrate and some players choose to display that as 64kbps.You are aware of selection effects? People who change codecs tend to reply, but people who don't do not. It not much use telling everyone that you use FLAC and will be using FLAC, but telling that and why you've moved to Wavpack can inform others.Then, whats the reason you expect a FLAC-exodus? The codec is just a very mature one, with no apparant reason to move away.Tak is impressive, but that doesn't mean people will hop over like sheep. I personally think it's not impressive enough to warrant a total conversion. Like you said, a 1-2% smaller library doesn't really matter and with todays processors it your hardrive that limits encoding/decoding speeds.FLAC has the best support any lossless codec has. There are definitly a large set of DAP that support it, every server-like appliance does, Denon/Marantz is moving to support FLAC in their CD/DVD players. If that's not al least an indication it's getting critical mass I don't know what is.

Whether TAK gains the following that it perhaps deserves or not many lossless users should be thanking Thomas for the work that he has put in.

It appears to me the the interest in TAK has helped to ensure that other codecs continue to improve. As Josh has already stated on various occassions, when the source code is released it will no doubt help them further.

Everyone's a winner baby.

Can we drop the pissing contest please?

ZOMG TAK!!!1!

I'm impressed by Tak's technical merits, but of course it takes more than technical merits to make a good lossless codec. FLAC is a portable project under a BSD license, and Josh works his buns off to garner hardware support from OEMs. And it's taken years to get that way.

I say hurrah to Thomas, hurrah to Josh, hurrah to Bryant, and warn that whatever happens with "mass migrations," it sure as hell won't be *soon*.

Hear here!!

I didn't know a tak development discussion would devolve into that. Seriously...

New presets in V1.0.2, Update

And another (final) update of the new preset system in TAK 1.0.2.

More changes:

1) I have sacrificed a tiny bit more compression for faster encoding.

2) The strongest compression mode of 1.0.1 (-p4m) is back as -p5m. It can show you, that the higher speed of the new version has not only been caused by a new preset configuration, but also by code improvements: It's now 62 percent faster without loosing compression!


Here comes a table comparing compression and encoding speed of V1.0.2 and V1.0.1:

Test system: Celeron Tualatin (P3) 1200 Mhz
Program version: V1.0.2 eval 07
File set: rw (46 files))
Speed measurements without disk io.


Because -p5 and -p5m were not available in V1.0.1, i compared them with -p4 respectively -p4m.

And here a comparison of the compression of TAK Turbo, Normal and the maximum setting (-p5m) with FLAC -5 and -8:


The work on TAK 1.0.2 might take about 1 to 2 more weeks.

Thomas

The new presets look quite promising, in my eyes it actually pays out that you made these changes to the predictors. The losses in compression are negligible, while the gains in processing speed definitely aren't. This is especially valid for p1 up to p4. I'm not really familiar with the performance of these presets, since Synthetic Soul's Lossless Codec Comparison doesn't include them, but the p5m one still looks like a good one for archiving to me. On Windows systems TAK should've become a decent archiving codec by now, being supported by widely available applications like Winamp (via your plugin), foobar2000 and MP3tag.

Tak 1.0.2 is looking even better with the revised presets.

TBeck, in what version should we expect piping support in TAK?

I'm a happy TAK p4m user, avg bitrate on my collection: 770 kbps.

I can't wait to get my hands on the -p5m.