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: faac 1.29* (Read 9861 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

faac 1.29*

hello! is anyone can do ABX-test of FAAC v1.29.9.2? or, where is i (we, those interested) can read about this?

Re: faac 1.29*

Reply #1
As far as I know, there are no listening tests of FAAC v1.29.9.2, but I may be wrong.

 

Re: faac 1.29*

Reply #2
There are some that involved FAAC as a low anchor because of its known poor quality.

If you need an AAC codec, I suggest FDK-AAC.


Re: faac 1.29*

Reply #4
If you want some information, you can follow the thread where some 1.29.x improvements where done and explained:
https://hydrogenaud.io/index.php?topic=114363.0

Basically, you can expect it to be much faster and a bit better in quality than 1.28, but fdk-aac and Apple AAC are two top-notch encoders.

Re: faac 1.29*

Reply #5
If you want some information, you can follow the thread where some 1.29.x improvements where done and explained:
https://hydrogenaud.io/index.php?topic=114363.0

Basically, you can expect it to be much faster and a bit better in quality than 1.28, but fdk-aac and Apple AAC are two top-notch encoders.

The most interesting commits I saw in FAAC semi-recently were replacing some old ISO MPEG code that was non-Free with some files under the GPL. Obviously, replacing non-Free software with Free software is good as it moves FAAC towards being less of a set of patches against proprietary software with GPL incompatible licenses (thus, no legal way to ship an FAAC binary*) towards an actual encoder.

I doubt the situation where LAME MP3 sounds leaps and bounds better than an AAC file produced by FAAC has changed substantially as the gap was just too wide before.

From what I've been able to gather, quality wise it's Apple>FDK>Nero>FFMpeg Native>FAAC.

If you're only interested in Free Software, then FDK>FFMpeg Native. So not a lot of options, and when you consider that VBR in FFMpeg doesn't actually produce usable results, then FDK is definitely the way to go.

Not only is Apple's non-Free, you have to run additional proprietary software (Quicktime....iTunes...) that is only meant to work on Macs and Windows. I'm not sure what, exactly, this entails, but Apple has a long history of releasing really bad and bloated Windows software. At one point, it was so bad that it even caused Windows Vista SP2 to break and become impossible to start up. There were also various scandals such as drive-by downloading software that the user didn't ask for, and their horrendous record on security (the Safari carpet bombing incident.)

So, my recommendation for FDK stands. It is high quality, Free Software, and cross-platform. The only significant limitations that I was able to determine was no native support for using FLAC as input and no support for 24-bit PCM input. If those aren't issues, then have fun.

*(The GPL requires that the entirety of the covered work have the GPL applied with no further restrictions. This is fine with permissively licensed software such at MIT, BSD, etc., but the ISO MPEG code has restricted the use of the code to "conforming products" that "implement the AAC standard". As this is not only a further restriction, but neither meets the FSF Free Software  nor the OSI Open Source Definition, there is no legal way to ship a binary of FAAC. Furthermore, the non-Free license in the ISO demo code warns that AAC is encumbered by software patents and that you have no rights to those patents, even though some of the companies that wrote the software hold patents. So it's a "Here's some software and we can sue you if you use it." license on top of everything else.)

Re: faac 1.29*

Reply #6
@DaemonFC : I assume you were adding to my response, not contradicting me, since we said the same.
(btw, before the ffmpeg native encoder, they had an even worse encoder. visualOn)


Out of curiosity (I didn't try this back then), I started a small quality test, and so far, it doesn't look nice at all.
1.29 from github is giving me lower audio quality at q 100 and at b 128 than 1.28 from rarewares on several files that I've tried to encode.

I observed that at q 100 it generally encodes at a lower bitrate than 1.28, and at -b128 it encodes at a higher bitrate than 1.28.
(Higher bitrate on same -b might be because it uses a higher lowpass, but the performance on -q is quite unexpected).
Also, on one of the tracks, even when using a higher bitrate than 1.28, it had more noise (maybe a bug) on a ressonant bass track.

I will try to prepare a few samples if anyone finds it worth.

Re: faac 1.29*

Reply #7
faac has come a long way, and does a better job today with knik's improvements. It's biggest advantage is encoding speed. At quality it's still inefficient compared to the top competitors, and misses things even at a quality of 1900. but unlike other encoders faac 1.29.9 won't cap out at higher rates. Q values of 4000 or higher typically producing 480-600 kbps may not be very useful to everyone, but it does fill the gap between lossy and virtually lossless.

FDK caps out easily in how many bits it allocates per channel, and begins to pad after that. On top of that, it doesn't do VBR well (granted, early implementation). But it is good for CBR, and typically renders better side fidelity than Apple (a "clearer" stereophonic sound).

(Many personal listening tests and spectral analysis used to form my opinions.)

Re: faac 1.29*

Reply #8
FDK [...] is good for CBR, and typically renders better side fidelity than Apple (a "clearer" stereophonic sound).

(Many personal listening tests and spectral analysis used to form my opinions.)

Can you elaborate the "side fidelity" part? Will increase of bitrate help?

Re: faac 1.29*

Reply #9
Apple's AAC encoder is great at covering the center, the common area where the left and right channels have the same information. The problem is, it spends too many bits filling the center spectrum, and not enough on side separation, where the left and right differ. It affects overall stereo fidelity and makes the sound less "clear" compared to the original.

Hearing this loss directly doesn't even require subtracting right from left in Audacity. It could be heard plugging in an aux jack incorrectly.

Raising the nominal bit rate does help. But unfortunately, even at 320 kbps, you can get effects like this:



Compared to FDK:



Track encoded: The Curse by Audioslave. The isolated sequence of the Apple encoding above sounds awful. Chris Cornell's voice there starts to sound like shingles clanging.

Re: faac 1.29*

Reply #10
Interesting! I too observed the empty patches in the spectograms but I always thought QAAC is being very efficient and removing inaudible stuff.

If QAAC has such an issue why is it considered the #1 AAC encoder?

Re: faac 1.29*

Reply #11
Interesting! I too observed the empty patches in the spectograms but I always thought QAAC is being very efficient and removing inaudible stuff.

If QAAC has such an issue why is it considered the #1 AAC encoder?

All encoders prioritize the center channel over the side channel, so it isn't necessarily a bad thing to see QAAC (or any encoder) doing that.  It is only bad if it creates an audible artifact. 

Re: faac 1.29*

Reply #12
To get good sound (not transparent) I arrived at -q 80 -c12000 (50-100k vbr music / speech) , -b 100 -c12000 (abr music) , -b 50 -c12000 (abr speech)