I recall there were some comparisons in the past that showed AAC codecs are inferior to MP3 in terms of audio quaility even at low bitrates (correct me if I'm wrong). Has anything changed since then? Also I wonder if there're some newer tests of Ogg Vorbis vs HE-AAC @ 64,96,128 and 192Kbit bitrates.

Mp3pro has been fairly close to old he-aac implementations, but not anymore. mp3 is far worse at low bitrates compared to lc-aac, let alone he-aac.

OGG & AAC (not he-aac as this is only for low bitrates) have been tied in the latest ~128kbps tests.

Unless you're comparing a very bad AAC codec to a good MP3 encoder, the answer is simply: not a chance in hell



HE-AAC outperforms Vorbis at low bitrates (<=64), normal AAC loses to Vorbis at low bitrates (<=64) and AAC ties at the higher end (>=128kbps). There are no recent tests at the bitrates inbetween (by more than 1 person).

"far" worse? I'd like to see that proven.



I'm sad to say I have to agree here (being a vorbis fanboy)... but for the 64-128kbps range...
http://forum.hardware.fr/hardwarefr/VideoS...0-1.htm#t922011
and
http://www.hydrogenaudio.org/forums/index....ndpost&p=312296

vorbis wins

24kbps . default samplerate for: MP3 -> 11Khz mono. LC-AAC -> 16Khz stereo, 22Khz mono, HE-AACv2 44Khz Parametric-stereo.

I think that this is enough but you can try for yourself. Or what does "low bitrates" mean to you?

he-aac has to be compared to mp3pro not regular mp3

CT lc-aac 24kbps 22khz stero -- default lowpass at just over 5khz (from winamp 5.2)
lame(3.93.1 --nspsytune or 3.97b2) -h --lowpass 5.5 --resample 12 -b 24 (stereo)
where is the lowpass set on nero's lc-aac at 24kbps? (i dont have nero so i dont know)
itunes lc-aac @24kbps i dont know either...

samplerate is irrelevent, only bitrate and audio bandwidth matter

mp3 'far' worse than lc-aac don't make me laugh (too late btw)
but then i guess it also depends on your definition of 'far'

Which part of "by more than 1 person" did you not understand?

Read for example
http://www.hydrogenaudio.org/forums/index....ndpost&p=357953

And understand why one should be sceptical of such tests. I think DSBR would be interesting to test at this bitrate, too.

Why then noone wants to fill the gap and run such comparisons? Is it so difficult/time consuming? I'm only asking that because my interest here is to find the best lossy codec at 48-96Kbit/sec bitrate (stereo input).

At least somewhat outdated 64KBit tests are here. The result of the tests is disappointing

Then why not providing results regarding the current 48kbps AAC test?

Birdie... This thread is here to contribute to the community, not whine...

HA has been running a lot of audio tests lately. Maybe a 80kbps listening test would be nice (but it will be after the multiformat 48, coming right up after this one.)

gameplaya15143:

What's the reason of wanting to cause controversy?

default samplerate for 24kbps CBR:

Lame 3.97b2, with "-b24" (stereo): 8Khz ( banwidth almost 4Khz). (resampling and increasing the bandwidth via custom switches will only increase artifacts)
Lame 3.97b2, with "-b24 -m m" (mono): 16Khz ( banwidth almost 6Khz). (I was wrong here, but the bandwidth is almost that of a 11Khz signal, and (11khz) is the usual samplerate in internet streams of this bitrate)
LC-AAC from Nero 6.6 (stereo): 22Khz (bandwidth almost 5Khz ). (So, a bit lower than what i said, but not much more)
LC-AAC from Nero 6.6 (mono): 44Khz (bandwidth almost 10Khz). (In practice, like what i said)
Vorbis (aotuv b4.51) (stereo) : 16Khz ( bandwidth 6,5Khz ). (Sounds quasi-mono)
Vorbis (aotuv b4.51) (mono) : 22Khz ( bandwidth 10Khz ).


MP3Pro from Nero 6.6 (stereo): 32Khz (bandwidth almost 16Khz). (In practice, it is almost mono).
HE-AAC (v1) from Nero 6.6 (stereo): 44Khz (bandwidth 11Khz with some spikes up to 16Khz but with dropouts) .
HE-AAC (v2) from Winamp 5.2 (stereo): 44Khz (bandwidth 16Khz).


MP3 vs LC-AAC : Increased bandwidth ( from better efficiency ). Quality can only be judged with a listening test.

MP3Pro vs HE-AACv2 : Similar bandwidth. Quality can only be judged with a listening test.


Edit: Added Ogg Vorbis for completeness.

I'm not sure if Lame is the best demonstration regarding mp3 @24kbps.

too many Xiph Zealots (not that I am one or anything ). No, but I would stream with both of them I like both of them equally.

did you even try it?
well.. since you like defaults so much... lame -b 32 .. bandwidth ~5.5khz with 3.97 (5khz with 3.93)

so is 8kbps a 'huge' difference? does an 8kbps difference mean that aac is 'far' better than mp3?..... 'slightly' maybe, but not 'far'

so when does the 24kbps listening test start I can't wait

... and now something on topic.... dualrate he-aac (44.1/88.2) in the 96kbps range would be interesting to test (but my vote still goes for vorbis )
what would you suggest then?

Why not use Lame as a representative? I don't believe any other mp3 codec would preform better than it.

At such low bitrates it is likely that FhG would perform better. Lame is not optimized yet for such bitrates.