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.
Recent Posts
51
Lossless / Other Codecs / Re: New lossless audio codec in development
Last post by Hakan Abbas -
Quote from: Porcus on
What does it do, really? Sure fixed predictors, then a Levinson-Durbin on who knows how you have (or have not) windowed the data - and then: Does the machine learning start to learn from there on, or from scratch?

And are those coefficients at the end really ... did a clever third-order really improve that much over a different least squares algorithm? It shouldn't ... should it? 
DLP is completely different from Levinson-Durbin. However, it can be thought of as a dynamic version of fixed estimators. The learning mechanism in DLP is actually trying to find the best case for a block selected from zero (e.g. 512 samples), i.e. the case with the least error. Here a decision is made by looking at only 2 or 3 samples. And the mistakes made are tried to be improved. This is not very interesting.

But what is interesting is that once the appropriate parameters are set, the best result can be obtained with the same parameters in the previous or subsequent blocks. Even if this depends on the shape of the data, in my tests it can sometimes be valid for hundreds of blocks before or after. you can see this by trying it immediately. Maybe we will also see the negative aspects of the method.
52
Lossless / Other Codecs / Re: New lossless audio codec in development
Last post by Porcus -
What does it do, really? Sure fixed predictors, then a Levinson-Durbin on who knows how you have (or have not) windowed the data - and then: Does the machine learning start to learn from there on, or from scratch?

And are those coefficients at the end really ... did a clever third-order really improve that much over a different least squares algorithm? It shouldn't ... should it? 
53
AAC - General / Re: QAAC: discussion, questions, feature requests, etc.
Last post by nu774 -
You can download libsndfile (either win32 or win64) from here: https://github.com/libsndfile/libsndfile/releases.
In the zip file you only need sndfile.dll under the bin directory, which you can copy to the same directory as qaac.exe(or qaac64.exe).
After you copied the sndfile.dll, run qaac --check (or qaac64 --check) from the command prompt.
If libsndfile is recognized and loaded by qaac, something like "libsndfile-1.2.2" is printed.

Since sndfile.dll at the github is statically linked to libogg and libvorbis, you don't need additional libogg/libvorbis DLLs.
In other words, even if you want to try other libogg/libvorbis binaries, you can't let qaac to use them.
55
Lossless / Other Codecs / Re: New lossless audio codec in development
Last post by Hakan Abbas -
Quote from: mycroft on
I'm not arguing about linear audio modeling. I think this approach is superior to any one produced before.
My dream is to code very very efficient float hybrid (lossless+lossy) audio codec.
https://github.com/Hakan-Abbas/Dynamic-Linear-Prediction
You can try DLP. Maybe you can see what I can't. This is a new linear prediction method that I worked on a long time ago. It works much more efficiently, especially for small blocks. For larger blocks (over 4,000 samples) the efficiency may decrease.

Apart from this test tool, I haven't had time to measure its performance in a real application yet. The main reason for this is that the processing speed will decrease a bit. The current best parameters need to be updated again at regular intervals. And how long that interval is needs to be determined accurately. Because while it is efficient to go with the best for each small block, there will be a significant penalty in terms of processing speed. If I can integrate it into a real application, I plan to share how it works. I need to be completely sure of the results.

Here's how it works briefly. First a csv file(no, value) or a 16 bit, 2 channel wav file is loaded. If the data is more than 100,000 samples, no graph is drawn. To see the graph, the slider at the bottom should be narrowed by pulling the slider from the right and left to the desired range. For DLP training, samples in the range of 100-1000 are sufficient. The smallest error sum for the relevant block is manually determined by changing the parameters. Then, we can see that these parameters work equally well for many previous and next blocks(according to my observations). So with a quick experiment/training on a small dataset, we can create a very good predictor for a larger dataset. DLP gives better results than Levinson-Durbin in most cases. We can see this by testing, except for very large blocks and very complex data. In addition, we need to know only the 2-3 parameter obtained for DLP.

Various tests can be performed with fixed estimators, Levinson-Durbin or DLP. The parametric data on absolute error and squared error sums can be seen at the bottom.
56
Polls / Re: 2025 Lossy format poll
Last post by a.ok.in -
Btw there's already decoding support on FFMPEG (native and FDK) and in some media players (not VLC, but MPV, MPC-BE and PotPlayer at least).
57
General - (fb2k) / Album list view for single genre?
Last post by gorman -
If it is possible to do, what would be the syntax to have  an album list view set to show a single genre?

I normally browse by artist-->albums with this:

$if($stricmp($substr(%album artist%,1,4),'The '),$stripprefix(%album artist%)', The',%album artist%)|%album%|[[%discnumber%.]%tracknumber%. ][%track artist% - ]%title%

I would like to basically replicate this view, just limiting it to a single, given genre (in my case the genre would be 5.1).

I have tried reading the wiki but found few examples of this conditionals use... plus I'm dumb, quite likely. ;D
58
foobar2000 mobile / Re: Certain mp3-files break playback
Last post by AlexForFoo -
Alas, it doesn't. In fact, it's quite worse. Enabling this option causes clicks/crackings during playback in intervals of some seconds. (Again, this is only on the new phone, on older ones playback with the option enabled works flawless.)
60
CD Hardware/Software / Re: Bought the same CD twice, got THE SAME C2 errors twice!!!
Last post by Feilakas -
Quote from: .halverhahn on
Please try to copy the CD without applying the AccurateRip offset correction - "offset correction value = 0".

Decades ago I had some stange C2 errors with a CD copy - that turned out to be a "cd-recorder software issue" - So a manufacturing error is quite possible.
https://hydrogenaud.io/index.php/topic,26162.0/prev_next,prev.html#new
Yeah, the C2 errors are identical and both CD surfaces are immaculate so I cannot possibly imagine it being anything OTHER than manufacturing error in the master.
Quote from: DVDdoug on
Quote
CueTools can even "repair" it.
Maybe there was a good batch and a bad batch.   (I don't know that would give these results.)

Is it repaired?   No audible errors?
CueTools has three "repair" options, I tried all three and all of them produce tracks that "sound right" but don't "CRC check right".

In all cases the CDDB database finds it "accurate" but AccurateRip does not "no match".

All other tracks are always a match though so is it due to a differing offset? Dunno