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: MPEG-4 Audio Lossless: final specifications (Read 127894 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

MPEG-4 Audio Lossless: final specifications

Reply #100 –
They took nine months to give birth to the mp4 solution 

MPEG-4 Audio Lossless: final specifications

Reply #101 –
From what I gather, this encoder is ok for end-user, but not ok for distribution in a freeware or commercial product?

MPEG-4 Audio Lossless: final specifications

Reply #102 –
just a stupid lil question here... will lossless formats ever reach 5:1 on the source? or is that technically impossible to do?

MPEG-4 Audio Lossless: final specifications

Reply #103 –
Quote from: Bourne on
just a stupid lil question here... will lossless formats ever reach 5:1 on the source? or is that technically impossible to do?


"what source?"

MPEG-4 Audio Lossless: final specifications

Reply #104 –
Hi all. I am ewby on this forum. For me i s very interesting to read all those things you guys write. But to the moment i did not understand wich software do you using to encode wav,s to MPEG 4 losless. I am using foobar2000 to encode all my cd collection to WavPack. But i am wondering to try this new format too. But i do not know how  10x
Listening to music is the best pastime.


MPEG-4 Audio Lossless: final specifications

Reply #105 –
Quote from: Garf on
"what source?"


CD PCM for example.... FLAC does it 2:1.... will it ever reach 5:1 ? Or technically impossible?

MPEG-4 Audio Lossless: final specifications

Reply #106 –
Quote from: Bourne on
Quote from: Garf on

"what source?"


CD PCM for example.... FLAC does it 2:1.... will it ever reach 5:1 ? Or technically impossible?


FLAC does not compress 2:1 in the general case. It's more like 1.7:1, sometimes much more, sometimes almost nothing. That's why I said "what source"? These codecs do not have fixed compression ratios and vary a lot depending on music style.

MPEG-4 Audio Lossless: final specifications

Reply #107 –
On multichannel it might be possible in the future to achieve something like 5:1 lossless compression, but on 2 channels it seems quite unlikely (in the general case of course).

MPEG-4 Audio Lossless: final specifications

Reply #108 –
Quote from: Bourne on
Quote from: Garf on

"what source?"


CD PCM for example.... FLAC does it 2:1.... will it ever reach 5:1 ? Or technically impossible?


For lossless 5:1 (average) is likely impossible. There are several algorithms used for lossless, though the average compression ratio does not vary much between them. Usually 1.5-2.5:1

In theory it's possible for certain data sets to actually be enlarged when attempting to losslessly compress.

See:
http://www.faqs.org/faqs/compression-faq/p.../section-4.html

The entire compression FAQ is good, though a little dated. I don't think much has changed since it was created in 99:
http://www.faqs.org/faqs/compression-faq/part1/

Really I don't think there's much of a need as lossy (mp3) is now so good, and it gives an outstanding 10:1 compression ratio with little quality loss. Usually provides a good approximation of the original signal, at least to human ears. It's just a much more intelligent way to compress audio as it makes assumptions based on weaknesses in human perception (less senstive to high frequencies, masking effects) and it combines both a compact mathematical representation (MDCT) with huffman coding (additional compression). It also assumes you have a fast enough computer to transform things back to the time domain 'real time' - these days a safe assumption 

Still the lossless compression is an interesting problem. It would be nice to get another factor of two - say 4:1. Lately I've been wondering if there might be some way to trade processing power for a greater lossless compression ratio. It might be possible, but if it takes a day to compress a CD I doubt many people would want to use the program.

Somebody needs to think of a new approach.

MPEG-4 Audio Lossless: final specifications

Reply #110 –
Quote from: kurtnoise on
Ok, reference encoder with mp4 support is available now but not playable yet. Sweet......


Moreover Windows COMP is suggesting me that files producted by RM18 seems to be identical to RM17 ones...

Anyway, this encoder shows some uncommon behaviour. Without intentionally looking for limit conditions, in my standard 26 files corpus SetF I have:

Code: [Select]
MP4ALS    RM17        -7 -n2560 -o32        64,19%
LA        v0.4b       high                  65,99%
OptimFrog v4.600exp   ultranew experimental 66,16%
Monkey    v4.01       Insane                66,34%


on one of the samples, showing a "-o32" MP4ALS mode compressing 1,8% - 2,2% more than best LA, OFR and APE modes. Nice. On the other hand, on other two samples I have:

Code: [Select]
OptimFrog v4.600exp   fast                  55,05%   45,42%
Yalac     v0.11       fast                  55,80%   49,36%
Monkey    v4.01       fast                  58,01%   49,48%
WavPack   v44a3       -hh                   60,30%   49,83%
MP4ALS    RM17        -7 -p -t2             60,68%   50,09%


which shows best MP4ALS mode compressing 2% - 5% worse than OFR, YAA, APE fast modes and also slightly worse than WV -hh. I know that codec performance strongly depends on the set you're using (and that this particular encoder is said to be immature, even if this same behaviour is shared by its predecessor LPAC), but that is maybe far too much.

Every encoder has its high and low peaks but in my experience (and checking the various comparison pages around) MP4ALS performance variability looks much higher than every other modern codec. Nevertheless its peak performances are very interesting and I'm really hoping some of the talented guys around will give us a better version, someday...

That is not to talk about decoding speed: in the same set (at -7 -p -t2), slowest file decodes around 7 times slower than fastest one...

MPEG-4 Audio Lossless: final specifications

Reply #111 –
Quote from: Josef Pohm on
which shows best MP4ALS mode compressing 2% - 5% worse than OFR, YAA, APE fast modes and also slightly worse than WV -hh.


Best mode should be -z3, not -7 -p -t2

MPEG-4 Audio Lossless: final specifications

Reply #112 –
Garf, do you still have the patch for some optimizations ?

MPEG-4 Audio Lossless: final specifications

Reply #113 –
Quote from: Josef Pohm on
Anyway, this encoder shows some uncommon behaviour. Without intentionally looking for limit conditions, in my standard 26 files corpus SetF I have:

Code: [Select]
MP4ALS    RM17        -7 -n2560 -o32        64,19%
LA        v0.4b       high                  65,99%
OptimFrog v4.600exp   ultranew experimental 66,16%
Monkey    v4.01       Insane                66,34%


on one of the samples, showing a "-o32" MP4ALS mode compressing 1,8% - 2,2% more than best LA, OFR and APE modes. Nice.

Josef, I think this is a rather extraordinary result! Would it be possible for me to get this sample to see if I can find anything special about it?

Thanks in advance.

David

MPEG-4 Audio Lossless: final specifications

Reply #114 –
Quote from: kurtnoise on
Garf, do you still have the patch for some optimizations ?




No, this format isn't going anywhere anyway.

MPEG-4 Audio Lossless: final specifications

Reply #115 –
Quote from: Garf on
Best mode should be -z3, not -7 -p -t2

You're obviously right, my mistake. When speed is not an issue '-z3' is a totally different story. Still a little slower, better compression, stable performance level.

And it shows no wonder:

Code: [Select]
MP4ALS    RM17        -7 -n2560 -o32        64,19%
LA        v0.4b       high                  65,99%
OptimFrog v4.600exp   ultranew experimental 66,16%
Monkey    v4.01       Insane                66,34%
MP4ALS    RM17        -z3 -p -t2 -b         66,42%

and no shame:

Code: [Select]
MP4ALS    RM17        -z3 -p -t2 -b         53,69%   43,38%
OptimFrog v4.600exp   fast                  55,05%   45,42%
Yalac     v0.11       fast                  55,80%   49,36%
Monkey    v4.01       fast                  58,01%   49,48%
WavPack   v44a3       -hh                   60,30%   49,83%
MP4ALS    RM17        -7 -p -t2             60,68%   50,09%

That said, in my opinion things do not change that much. It still looks a little odd to see a '-o32' mode over 2% better than a '-z3' mode on the first sample and an almost 7% gap between '-z3' and '-7' on the other ones.

Quote from: bryant on
...would it be possible for me to get this sample to see if I can find anything special about it?

Of course! I hope you got my PM. Though currently I'm not done with it, on the same sample the second best encoder is... RkAu!
 

MPEG-4 Audio Lossless: final specifications

Reply #116 –
I said -z3, not -z3 -p -t2 -b. These settings might not even work correctly.



ALS is really 2 totally different encoders:



1) foward prediction like FLAC: used by normal settings, and -7 and so, with extras like -p -t2 -b etc



2) backwards prediction like APE and WavPack: used by -z settings, won't combine with most of the other methods



You can use on or the other, but not both.

MPEG-4 Audio Lossless: final specifications

Reply #117 –
How does mp4 als relate to AAC-HD (which Fraunhoffer are promoting)?

MPEG-4 Audio Lossless: final specifications

Reply #118 –
mpeg4als RM18 released 

uri: http://www.nue.tu-berlin.de/research/proje....html#downloads

-MP4 option implemented, but can't play at present 

Code: [Select]
"test.mp4":
Track   Type    Info
1       audio   MPEG-4 Unknown Profile(31), 343.066 secs, 440 kbps, 44100 Hz
<name>madoka</name>

MPEG-4 Audio Lossless: final specifications

Reply #119 –
Quote from: spoon on
How does mp4 als relate to AAC-HD (which Fraunhoffer are promoting)?


Got a link? Never heard of it.

MPEG-4 Audio Lossless: final specifications

Reply #122 –
So "HD-AAC" is really MPEG-4 SLS. It has nothing in common with ALS except that they can both operate in lossless mode.

MPEG-4 Audio Lossless: final specifications

Reply #123 –
Quote from: madoka@ex-sounds on
mpeg4als RM18 released 

uri: http://www.nue.tu-berlin.de/research/proje....html#downloads

-MP4 option implemented, but can't play at present 

Code: [Select]
"test.mp4":
Track   Type    Info
1       audio   MPEG-4 Unknown Profile(31), 343.066 secs, 440 kbps, 44100 Hz


I submitted some changes to let MPEG-4 ALS show correctly in tools built on MPEG4IP.  Bill checked them into CVS, but they aren't in an official release yet.  Still can't play them though

MPEG-4 Audio Lossless: final specifications

Reply #124 –
Quote from: benski on
I submitted some changes to let MPEG-4 ALS show correctly in tools built on MPEG4IP.  Bill checked them into CVS, but they aren't in an official release yet.  Still can't play them though

Wow!! good news!! thx benski   

I'm waiting for an official release of new winamp or fb2k 
<name>madoka</name>