2005-12-30, 02:03 | Link #1 | ||
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
The AAC discussion thread (split from AVC)
Quote:
I hav a same problem with lamune episode1, too, the video is h264, audio is 64kbps aac.... IIRC nero HE aac encoder is outdated(new version will be availabe at nero 7++ verison, IIRC). and the quality is much worst than 3GPP reference enhanced-HE AAC-v2 encoder. Quote:
Ogg vorbis quality was improved alot since a guys called "aoyumi" made alot of tuning with this audio codec, from Aotuvb2 till Aotuvb 4.5.1. Last edited by SSS; 2005-12-30 at 02:26. |
||
2005-12-30, 04:16 | Link #2 |
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
Here is the link to download the 3GPP refernce encoder:
http://www.aziendeassociate.it/cd.asp?dir=/Misc the binary is "EnhAACPlus_630_ICL7(9).7z", ICL7 or ICL9 compiled. and the main webpage: http://celticdruid.no-ip.com/xvid/ |
2005-12-30, 04:33 | Link #3 | |
Translator, Producer
Join Date: Nov 2003
Location: Tokyo, Japan
Age: 44
|
Quote:
I've never heard 48 Kbp/s sound that good... I am totally using this for all future encodes, it DOES beat nero, hands down.
__________________
|
|
2005-12-30, 04:54 | Link #4 | |
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
Quote:
this encoder automatically switch to Parametric stereo when encode at 48kbps, higher than 48kbps is SBR only. Again, here is the link to download the Helix CLI Producder, it have he/aac encoder too(but not sure it can produce he-aac v2 or not): https://helixcommunity.org/download....11_windows.zip You have to register and agree the BEULA in order to download it. BTW helix community have a MP3 encoder too, more infos read here: http://www.hydrogenaudio.org/forums/...howtopic=35531 and here: http://www.hydrogenaudio.org/forums/...=35531&st=175# |
|
2005-12-30, 06:30 | Link #5 |
Rozen Detective
Join Date: Dec 2005
Location: Germany
Age: 40
|
Also, there is a listening test, which shows quite good just how bad the Nero encoder performs compared to other HE-AAC encoders. The Helix one seems to be the best (the reference encoder was not tested, I think).
|
2005-12-30, 13:49 | Link #7 |
Panda Herder
Join Date: Dec 2005
Location: A bombed out building in Beruit.
|
Some of the following threads may be of intrest:
http://forum.doom9.org/showthread.ph...listening+test (A lot of potentailly usefull links) http://www.hydrogenaudio.org/forums/...howtopic=32440 (deals with rencoding from a lossy source.) http://home.broadpark.no/~clarkt/Codec_Compare.pdf (A correlation based test, not quite sure what to make of it.) I definately need to take a look at the refernce encoder. Edit: I just tried the reference encoder. I think it's the first time I've heard 64kbps AAC that didn't make me want to poke out my eardrums. Last edited by DryFire; 2005-12-30 at 14:02. |
2005-12-31, 04:51 | Link #8 | |
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
Quote:
If ppl really need to encode 5.1 channel + VBR + low bitrate, better choose Ogg vorbis(though there still some problem with Ogg vorbis encoder/decoder ect ect are waiting to solve)... (Im sorry all about this off-topic, the off-topic stater is me , if can, Mod please move/split the off-topic to new thread...) |
|
2005-12-31, 05:17 | Link #9 |
Excessively jovial fellow
Join Date: Dec 2005
Location: ISDB-T
Age: 37
|
Vorbis is horrible for 5.1 stuff due to atrociously inefficient channel coupling. LC-AAC sounds better and is smaller... Additionally, 5.1 vorbis tends to upset decoders, even CoreVorbis, which doesn't even have a downmixing option. Vorbis is really great for stereo at modest bitrates, but for 5.1 it loses big time to AAC.
__________________
|
2005-12-31, 08:56 | Link #10 | |
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
Quote:
Here is the link to find some latest audio codecs info, including ogg vorbis. http://www.hydrogenaudio.org/forums/...act=SF&s=&f=57 Now ogg vorbis can handle multichannel correctly. Regarding the ogg vorbis decoder downmix problem, isnt ffdshow can do that?(but dont use libavcodec to decode ogg vorbis). |
|
2006-01-01, 11:05 | Link #11 |
Excessively jovial fellow
Join Date: Dec 2005
Location: ISDB-T
Age: 37
|
I tested it with libvorbis 1.1.2 built on 8th december 2005 recently, and while it wasn't as bad as it had been, it still didn't beat even LC-AAC. And ffdshow does NOT handle 5.1 vorbis correctly according to my (very brief) tests. CCCP staff has been trying to convince milan to include libvorbis as an alternative to libavcodec for vorbis decoding in ffdshow, because of weird libavcodec issues, but so far no success. He's not completely opposed to the idea AFAIK, but it's a low-priority thing...
__________________
|
2006-01-01, 23:51 | Link #12 | ||||
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
Quote:
Here is the free 5.1 ac3 sample i used to testing: http://www.lynnemusic.com/surround.html but i have one audio player(foobar200k), its only have problem to transcode 5.1ac3 to 5.1 ogg vorbis, that player created a low quality 5.1 ogg vorbis file(only LF/LR channel have quality problem IIRC, the problem still not clear). But quality bug will gone if use alternate way to transcode, like: ac3 -> wav/(lossless codec that can handle multichannel) -> ogg vorbis. I need another apps that can decode 5.1 ac3 to 6ch wav(waveformatextensible?/loseeless?), to confirm its either an Ogg vorbis encoder or decoder problem, but so far i havent found any.... Quote:
or you can try decode with corevorbis -> ffdshow(downmix), i will try it later. Quote:
yes, ffdshow libavcodec have seroius quality probem when decodig 5.1 ogg vorbis, termor doesnt have problem but its compiled with low accuracy... Quote:
How sad, ogg Vorbis doesnt have channel coupling |
||||
2006-01-02, 04:39 | Link #13 | |
Excessively jovial fellow
Join Date: Dec 2005
Location: ISDB-T
Age: 37
|
I used BeSweet GUI to transcode from AC3. I can play the result in Winamp...
Quote:
__________________
|
|
2006-01-02, 07:01 | Link #14 |
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
Can you upload the sample(30sec would be enough) original ac3 file, lc-aac encoded file and ogg vorbis encoded file. Im really interest how 5.1 aac quality compare with 5.1 ogg vorbis.
In my quick test, both using 64kbps VBR, 5.1 aac(Nero lc-aac encoder) indeed smaller than 5.1 vorbis, but i cant say its better than vorbis. Same thing happend with 2ch encoding. I can upload my source and encoded file too, if you guys are interested. |
2006-01-02, 10:53 | Link #15 |
Excessively jovial fellow
Join Date: Dec 2005
Location: ISDB-T
Age: 37
|
Sure. AC3 original withheld until someone less lazy and better with audio tools than I am tells me how to cut an AC3 without reencoding it.
The samples are taken from a japanese war/naval movie (Lorelei), since what we normally do isn't music, it's movie soundtracks. 60 seconds Ogg Vorbis sample. (2.7 MB) Q4, channel order FL, C, FR, SL, SR, LFE (this is the besweet default and what Winamp's Vorbis decoder claims is the right order). Encoded with libvorbis.dll using libvorbis 1.1.2, downloaded from rarewares.org (this one: http://www.rarewares.org/files/ogg/libvorbis1.1.2.zip) and BeSweetGUI 0.7b8/BeSweet 1.5b30. BeSweet commandline: Code:
"C:\Program\BeSweet\BeSweet.exe" -core( -input "c:\lorelei\VTS_02_1 - 0x80 - Audio - AC3 - 6ch - 48kHz - DRC - Japanese - DELAY 0ms.ac3" -output "c:\lorelei\6ch.Vorbis.Q4.ogg" ) -azid( --maximize ) -split( -start 570 -end 630 ) -ogg( -q 0.400 -6ch 1 ) Code:
"C:\Program\BeSweet\BeSweet.exe" -core( -input "c:\lorelei\VTS_02_1 - 0x80 - Audio - AC3 - 6ch - 48kHz - DRC - Japanese - DELAY 0ms.ac3" -output "c:\lorelei\6ch.HE-AAC.VBR-streaming.mp4" ) -azid( --maximize ) -bsn( -6chnew -vbr_streaming -codecquality_high -aacprofile_he ) -split( -start 570 -end 630 ) Code:
"C:\Program\BeSweet\BeSweet.exe" -core( -input "c:\lorelei\VTS_02_1 - 0x80 - Audio - AC3 - 6ch - 48kHz - DRC - Japanese - DELAY 0ms.ac3" -output "c:\lorelei\6ch.LC-AAC.VBR-internet.mp4" ) -azid( --maximize ) -bsn( -6chnew -vbr_internet -codecquality_high -aacprofile_lc ) -split( -start 570 -end 630 ) EDIT: I redid the tests with the aoTuVb4.51 vorbis DLL's, just for reference. Personally I can't hear much difference, but then again I don't even have a 5.1 system... 60 seconds Ogg Vorbis aoTuVb4.51 sample. (2.78 MB) Exactly identical to the first Vorbis one, except for the encoding library used.
__________________
Last edited by TheFluff; 2006-01-02 at 11:33. |
2006-01-03, 05:02 | Link #16 |
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
Thx for the sample files
Too bad, i cant heard any difference, or audio artifact from these 4 5.1ch audio files(may be this movie soundtrack is less complex than music..., or my ears have problem?) But judging from file size, lc/he aac is smaller. May be Ogg vorbis isnt design for 5.1ch encoding Now lets compare ogg vorbis with nero lc-aac, 2ch encoding: The source is Shuffle ED, Innocence(lossless). http://www.geocities.com/y0ngc/innocence-AAC.zip Encoder Settings: VBR: tape lowest, Encoder quality: high, LC-AAC, PNS, encoder version=4.2.1.0 http://www.geocities.com/y0ngc/innocence-BlackSword.zip Encoder settings: -q 0, Black Sword optimized Ogg vorbis encoder(based on AoTuVb4.51) I cant test with itunes lc-aac with 64kbps because stereo mode is not allow in such low bitrate IMO AoTuV4.51 is much better, has less audio artifact than nero lc-aac. |
2006-01-03, 06:31 | Link #17 |
Excessively jovial fellow
Join Date: Dec 2005
Location: ISDB-T
Age: 37
|
If you're going to encode at that low bitrates, not using HE-AAC is pretty weird. LC-AAC is made to sound good at higher bitrates, while HE-AAC is made to compress like mad while still sounding OK. The encoders even restrict usage of HE-AAC at higher bitrates, in favor of LC...
__________________
|
2006-01-03, 08:31 | Link #18 | |
Suki Suki Sia-chan ^Q^
Join Date: Dec 2005
|
Quote:
"CD quality at 48kbps" bold statement Ya im wired ppl I think this 64kbps listening test isnt fair, to Nero aac codec IIRC someone in hydrogenaudio.org said, he-aac isnt designed, or its useless to encode audio at high bitrate. (this ppl is "Garf", he is Nero mpeg4/aac codec Engineer, i couldnt remember what exactly he said...) If you want to encode he-aac at 128kbps(only), use Helix DNA producer |
|
|
|