Making a two part encode, need some guidance

[Chainmax]

I am about to try an encode an episode from one of my DVDs. It will be encoded with both the XBOX lowres and iPod profiles. Thing is, since the intro has so much motion, it will be encoded with a higher average bitrate. So, here are the questions:

1) What is the maximum bitrate for the XBOX lowres profile that ensures smooth playback? The one for the iPod is 768kbps, right?

2) For the XBOX version I'll be using Vorbis audio and MKV container. What tool should I use to join the MKV with the intro and the MKV with the rest of the episode?

3) I want the audio in the iPod version to the HE-AAC v2 @ 64kbps, which encoder should I use? And how do I join the two parts after they are done?

[Audionut]

My question to you is! Why make it a 2 part encode? Why not just use zones?

For starters, that would cancel out Q2 and Q3b.

edit: For Q1. I can say that sharktooth's ipod profile default's at 700kbps. And every other profile "including xbox lowres" default's at 1000kbps. So I think it's safe to say, atleast 1000kbps. Although i'm sure someone who know's more about xbox's can comment better on this.

For Q3a. This 48kbps AAC public test (http://www.mp3-tech.org/content/?48kbps%20AAC%20public%20test) is scheduled to end on the 19th of march. I suggest you wait for the results of that test.
http://www.hydrogenaudio.org/forums/index.php?showtopic=39731

[Hellworm]

Are you using 2 pass? There should be no reason to set different bitrates, every scene will consume as much as it needs for a constant quality.

[Chainmax]

Audionut: I can't believe I forgot about zones :o.

Hellworm: I'll be using 3-pass, but the intro requires 2.5x-3x as much bitrate as the episode itself.

[xyloy]

3) I want the audio in the iPod version to the HE-AAC v2 @ 64kbps, which encoder should I use? And how do I join the two parts after they are done?
MediaCoder (http://mediacoder.sourceforge.net/) uses Winamp's AAC plugins from Coding Technologies. And as reported in the AAC public test at hydrogen forums, it's as good as the real aac encoder(and always far better than Nero AAC).

MediaCoder is open-source, and can use 44.1 KHz PCM .WAV files for input. If you have an .AC3 or .DTS input, you can transcode it to .WAV with BeSweet.

PS: In MediaCoder, HE-AAC is only available at 48 Kbps or less. Maybe it's due to the MPEG-4 specs or the plugin... Anyway, the result is at least as good as 160 kbps Lame MP3. ;)

edit: to merge two audio files, you can try to do a copy/paste operation with Audacity(free and open-source) or SoundForge(not free at all :/ )
First open the second part, select all(Ctrl+A) => copy. Close it, open the first part, move the current selection to the end of the file and paste. But I only tested it with .wav or .mp3 files... :(
Sorry, I'm sure there is easier ways to do it, but I've only done it a few times.

[Sharktooth]

MediaCoder (http://mediacoder.sourceforge.net/) uses Winamp's AAC plugins from Coding Technologies. And as reported in the AAC public test at hydrogen forums, it's as good as the real aac encoder(and always far better than Nero AAC).

MediaCoder is open-source, and can use 44.1 KHz PCM .WAV files for input. If you have an .AC3 or .DTS input, you can transcod it to .WAV with BeSweet.

PS: In MediaCoder, HE-AAC is only available at 48 Kbps or less. Maybe it's due to the MPEG-4 specs or the plugin... Anyway, the result is at least as good as 160 kbps Lame MP3. ;)

edit: to merge two audio files, you can try to do a copy/paste operation with Audacity(free and open-source) or SoundForge(not free at all :/ )
First open the second part, select all(Ctrl+A) => copy. Close it, open the first part, move the current selection to the end of the file and paste. But I only tested it with .wav or .mp3 files... :(
Sorry, I'm sure there is easier ways to do it, but I've only done it a few times.
uhm... quite some misinformation here...
first, Nero 7 AAC+ encoder is as good as CT AAC+ encoder... difference is small and the HA test was made to estabilish what's best...
also AAC+ (with PS) is available at 48kbps or less... normal HE-AAC can go higher... and there's no way AAC+ at 48kbps can be as good as Lame.
Infact HA test shown 64kbps HE-AAC is not quite good as lame... and AAC+ at 48 is clearly inferior to HE-AAC...

[xyloy]

Nero 7 AAC+ encoder is as good as CT AAC+ encoder...
I was reffering to the old Nero AAC's encoder (back in 2003). I believed the test was the same as this (http://www.hydrogenaudio.org/forums/index.php?showtopic=36868&st=0) one. Oops! ;)

Infact HA test shown 64kbps HE-AAC is not quite good as lame... and AAC+ at 48 is clearly inferior to HE-AAC...
When I say 'HE-AAC', I mean always AAC used along with SBR and PS, that is what gives me very good results with CT's plugins at 48 Kbps(stereo, 44.1 KHz), unless my ears should be cleaned up. :D

But AAC+ is not the same as HE-AAC (http://en.wikipedia.org/wiki/AAC%2B) ?
Wikipedia says:
aacPlus was standardized by the MPEG under the High Efficiency AAC (HE-AAC) name.
If not, I'm becoming crazy.. >_<

Normal HE-AAC can go higher
So it's due to the plugin.
Thanx for the info, I'd like to test SBR's and PS's efficiency at higher bitrates(now the only thing missing for me is VBR..) when I'll have some time(and an encoder able to do it). ^^

[Chainmax]

What would you recommend me to use for 48 or 64kbps?

[shon3i]

Nero 7 AAC+ encoder is as good as CT AAC+ encoder Newer can be especialy at low bitrates and 6ch encode. btw Nero 6 is much better than Nero 7 and he is near quality to CT but is not better than CT. This is for HE-AAC and aacPlus v1/v2 encodings

and there's no way AAC+ at 48kbps can be as good as Lame.
You should be try.

[xyloy]

What would you recommend me to use for 48 or 64kbps?
Well, 64 kbps if you have enough space for it.
But 48 kbps for video+audio file when space is limited, especially when video bitrate is already low enough.

(IMHO)

[IgorC]

During yesterday I've made a little personal blind ABX test :

http://img145.imageshack.us/img145/6517/comparison9zo.jpg

As you see HE-AAC2 (sbr+ps) isn't better than HE-AAC1 (sbr) at 48 kbit/s.
As it says here http://www.codingtechnologies.com/products/aacPlus.htm

1. Nero SBR 3,67
2. Winamp SBR 3,29
3. Winamp SBR+PS 2,83
4. Nero SBR+PS 2,73

http://www.codingtechnologies.com/products/assets/3GPP_mushra_18_48kbps_2.gif

And HE-AAC at 48 kbit/s is worse than popular Lame 128 VBR 5 encoding. Even between excelent LC-AAC 96 kbit/s Itunes QT7 and Lame 128 VBR 5 I would still prefer Lame. But 48 kbit/s SBR for movie track is ok.

[Chainmax]

So, what AAC flavor and encoder would you recommend me at 64kbps?

[IgorC]

Instead of "bla-bla-bla" about how good can be HE-AAC at 64-80 kbit/s comparing to Lame VBR 128. Just visit http://www.hydrogenaudio.org/forums/index.php?showtopic=35438 and see comparison

http://audiotests.free.fr/tests/2005.07/80/80TEST_PLOTS_06.png
As you see Lame VBR 128 is still better than HE & LC-AAC at 80 kbit/s. I'm prety sure that Lame 128 is even better than itunes LC-AAC 96 kbit/s.
Since then AAC was impoved, but Lame also.

[guada 2]

Hello IgorC, :)

Please, can you specify the version of LAME used ?

Bye.

[xyloy]

But 48 kbit/s SBR for movie track is ok.
Of course. I'm very impressed by CT's HE-AAC sound quality at 48 Kbps, bot not enough to forget that 48 Kbps is a very low bitrate and start to use it for audio CDs ripping(even if that sort of thinking about audio bitrate is 'scaled' on my personnal Lame MP3 terrible experience :P ).

So, what AAC flavor and encoder would you recommend me at 64kbps?
You can try MediaCoder's AAC+ at 64 Kbps with "Stereo Independant" channel mode. ;)

PS(to Sharktooth):
My mistake. With MediaCoder, it's not SBR that is not available with more than 48 Kbps, but Parametric Stereo.

[shon3i]

I'm very impressed by CT's HE-AAC sound quality at 48 KbpsAlso and 6ch encode for films i great too comparing to original ac3 sound is very good. Nero 6/7 with 6ch encode is very terrible at same bitrate as CT, but vbr modes is good but cost about 50mb more space for file.

@IgorC can some provide 6ch encoding compare betwen Nero and CT and think that nero SBR isn't better then CT. Can i listen this files and send my scores.

[Audionut]

As you see Lame VBR 128 is still better than HE & LC-AAC at 80 kbit/s. I'm prety sure that Lame 128 is even better than itunes LC-AAC 96 kbit/s.
Since then AAC was impoved, but Lame also.

Of course. It has more bitrate.:confused:
But as you can clearly see, at the same bitrate, the other encoders are better. Clerarly Vorbis.

[Chainmax]

Sharktooth, does enabling the no fast pskip option need a beefier system to decode than disabling it? If so, why? If not, could I use it for the XBOX encode?

[IgorC]

@IgorC can some provide 6ch encoding compare betwen Nero and CT and think that nero SBR isn't better then CT. Can i listen this files and send my scores.
khem.... I don't understand your english well. But it's not problem anyway. Let's clear some points :
1. do You ask me to do 5.1 multichannel test? I'm sorry I haven't 5.1 conected to PC. I do have 5.1 for TV but it's too small and low q. kit
2. You want to listen encoded files? Sure I will upload them (with all info's results *.txt generated files )tonight/tommorow if you are interested.
3. Send your scores? You can do your personal ABX test. google it and find soft for abx test. its special program for such purpose that mixed blindly for all samles : original, CT winamp, nero so finally you can't say which one is. Only you have to do is to put blindly the score that you think it desired. And only then see the *.txt file with final scores.
Of course some artefacts hearing training will be requeired.
For more information visit www.hydrogenaudio.org

There is an open for public to participate in 48 kbit/s Stereo HE-AAC test http://www.hydrogenaudio.org/forums/index.php?showtopic=41598

Everybody is welcome to help to provide realable results.

Guada You will find all information about Lame version in gived link.

[shon3i]

khem.... I don't understand your english well. But it's not problem anyway. Let's clear some points :
Whell sorry about that.
1. do You ask me to do 5.1 multichannel test? I'm sorry I haven't 5.1 conected to PC. I do have 5.1 for TV but it's too small and low q. kit
Yes i mean on that, but you not must have 5.1 system to compare original ac3 and nero6/7/CT AAC. Well very good test can be on stereo speakers and files dominixed to stereo or DPL.
3. Send your scores?I mean can i join in this 48 kbit/s Stereo HE-AAC test and listen encoded files and then vote to my favor.

[IgorC]

shon31 Sorry. I have no time. Read private message.

[guada 2]

Lame 3.97 bêta 2
:thanks: for the link IgorC. :)

Very interesting.....

Do exist actually a test multichannel: MP3 surround LAME 3.97b2 vs OGG Vorbis surround vs HE-AAC surround ?
I think that it would be useful to see what these codecs audios would give at this level.

No..

[shon3i]

That was be great:
Nero 6/7 vs CT vs OGG vs Aud-X at same bitrates

[Sharktooth]

Sharktooth, does enabling the no fast pskip option need a beefier system to decode than disabling it? If so, why? If not, could I use it for the XBOX encode?
no and yes.

[Chainmax]

What other features could I enable/raise and still achieve a playable 512x384 encode for the XBOX?

[bond]

there is no mp3 multichannel encoder

fhg and aud-x mp3 surround are not real multichannel, but stereo with surround upmix info

[Chainmax]

:confused:

[nm]

What other features could I enable/raise and still achieve a playable 512x384 encode for the XBOX?
You should be able to use all the features available for a 512x384 encode and normal bitrates. However, I haven't used Xbox myself, but I've done tests with MPlayer on similar hardware.

[guada 2]

there is no mp3 multichannel encoder

fhg and aud-x mp3 surround are not real multichannel, but stereo with surround upmix info


Hello Bond,

Thank you for these precisions, and this clarification on the terms misused by some which don't reflect always the truth.

Bye.

[Chainmax]

You should be able to use all the features available for a 512x384 encode and normal bitrates. However, I haven't used Xbox myself, but I've done tests with MPlayer on similar hardware.

Not quite so, Sharktooth already said that custom matrices are a no-go.

[nm]

Well, I wasn't thinking of XBMC restrictions, but the speed of libavcodec's H.264 decoder. And actually you could use CQMs on Xbox by installing Linux on it and using MPlayer directly ;)

[Chainmax]

It's not a matter of wether it can or not, it's a matter of wether it should or not. Apparently using custom matrices is too much for the XBOX's CPU.

[nm]

I don't think decoding a file encoded with a CQM is harder than decoding a file without one if average bitrate and resolution aren't changed. Just tested with mplayer --benchmark and I only got a 2% difference in the decoding times (and that could be within error of measurement).

[3dsnar]

there is no mp3 multichannel encoder
fhg and aud-x mp3 surround are not real multichannel, but stereo with surround upmix info
Bye.
Maybe some explanation would be useful.
Aud-X, mp3 surround, or HE-AAC 5.1 as well are the same in terms of general concept of generating 6 channels (5.1). I.e. they utilize the main audio stream and some additional info to reconstruct the total 5.1 panorama (to 'unmerge' the downmix).


Such unmerging can be very efficient in terms of energy error and phase error (take a look at our whitepaper), so in fact it is difficult to say that this is something that is not a really surround...
For example, if you would preserve a perfect differential signals (and call it upmix info), you would be able to perfectly restore the 5.1 panorama, but still the main audible part would be stereo.
But a perfect reconstruction is not necessary, while it is necessary to preserve the perceptually high quality.

-------------------------
Another example. If you would take (for example) 6 mp3 tracks, and encode each
with 32 kbps (completely shitty quality). Than put all of the separate mp3s in some imaginary container. Than use 6 separate mp3 decoding processes to decode therm back to 6 channel PCM, would it be more 'surround' than the listed three coding methods?
I think not, because each of the 6 signals would be less similar to the original 5.1 source 6 signals (in terms of perception, or objectivized energy/phase error measurements).
So I would say that the way you recreate the 5.1 sound is irrelevant.
The resulting quality is of importance only.