eac3to - audio conversion tool

[sshd]

The Machinist HD DVD:
3: h264/AVC, 1080p24 /1.001 (16:9) with pulldown flags

This title does not appear to have pulldown flags.
Video is destroyed when removing pulldown with h264info and plays fine at 23.976 fps without.

[tebasuna51]

Thanks for your new eac3to version.

You includes the -extensible parameter to output wav with the WAVE_FORMAT_EXTENSIBLE header (henceforth WFE).
I like this improvement but I have some comments:
- When use WFE header is mandatory a fix channel order.
- The most useful new data (maybe the unique) is the Channel Mask to specify the channels present in the audiodata.
- When is unknown the channels present in the source is recommended use a default Channel Mask based in the number of channels.
- When we know the source channels the channel mask must be set in accord.

1) ABOUT DEFAULTS

I think the default used in eac3to for 4 channels (0x000F = FL FR FC LF) must be changed for 0x0033 = FL FR BL BR (Quadro) more usual and compatible with Flac/Ogg multichannel.
A WFE wav with maskchannel 0x000F is rejected by flac.

The default used for 8 channel (0x00FF = FL FR FC LF BL BR FLC FRC) is not usual, maybe the default can be 0x063F = FL FR FC LF BL BR SL SR

When I try know the default for 7 channels using a 3/3.1 wav file:

eac3to v2.45
command line: "D:\Test\AudioN\eac3to\eac3to.exe" "E:\Test\7_61.wav" "E:\Test\z331.wav" -extensible
------------------------------------------------------------------------------
WAV, 6.1 channels, 0:00:20, 16 bits, 48khz
Doubling 7th channel...
Reading WAV...
Writing WAV...
Creating file "E:\Test\z331.wav"...
eac3to processing took 1 second.
Done.

For what "Doubling 7th channel..."?
Maybe the default for 7 channel can be: 0x013F = FL FR FC LF BL BR BC

2) ABOUT KNOW SOURCE AC3

At least with ac3 source we can obtain perfect WFE wav output:

acmod.lfe ac3 channels      Mask and MS channels ordered      Detect-MaskCh eac3to   libav remap
-------------------------  ---------------------------------  ---------------------  ------------
1  1/0.0  C                0x0004  FC                         1.0    0x0004  ok      not needed
1  1/0.1  C LFE            0x000C  FC LF                      1.1    0x0003  (1)(2)  not needed
2  2/0.0  L R              0x0003  FL FR                      2.0    0x0003  ok      not needed
2  2/0.1  L R LFE          0x000B  FL FR LF                   2.1    0x0007  (2)     not needed
4  2/1.0  L R S            0x0103  FL FR BC                   2/1    0x0007  (2)     not needed
4  2/1.1  L R S LFE        0x010B  FL FR LF BC                2/1.1  0x000F  (2)(3)  -0,1,3,2,4,5
6  2/2.0  L R SL SR        0x0033  FL FR BL BR                2/2    0x000F  (2)     not needed
6  2/2.1  L R SL SR LFE    0x003B  FL FR LF BL BR             2/2.1  0x0037  (2)(3)  -0,1,4,2,3,5
3  3/0.0  L C R            0x0007  FL FR FC                   3/0    0x0007     (3)  -0,2,1,3,4,5
3  3/0.1  L C R LFE        0x000F  FL FR FC LF                3/0.1  0x000F     (3)  -0,2,1,3,4,5
5  3/1.0  L C R S          0x0107  FL FR FC BC                3/1    0x000F  (2)(3)  -0,2,1,3,4,5
5  3/1.1  L C R S LFE      0x010F  FL FR FC LF BC             3/1.1  0x0037  (2)(3)  -0,2,1,4,3,5
7  3/2.0  L C R SL SR      0x0037  FL FR FC BL BR             5.0    0x0037     (3)  -0,2,1,3,4,5
7  3/2.1  L C R SL SR LFE  0x003F  FL FR FC LF BL BR          5.1    0x003F  ok      already done

(2) Default mask for channel number, the correct mask can be easyly put based in detection
(3) When decoded with libav need remapping channels.

EDIT: My previous note (1) was:
(1) Wrong detection, must be 1.1 (not important because mono + LFE is really strange).
But is my fault, eac3to detect 1.1 correctly

[calinb]

Originally Posted by madshi

At this point in time eac3to is VERY picky about the source material.

Indeed! Oftentimes, source that produces the "This track is not clean...Please clean the track with delaycut" error can be processed with eac3to ver. 1.16--after remuxing/joining the .evo(s) with EVOdemux, if necessary.

[gregt]

I seem to have run into a consistent problem with certain DVDs that are 480i60. The following is an example:
eac3to v2.45
command line: eac3to vts_01_1.vob+vts_01_2.vob+vts_01_3.vob+vts_01_4.vob t.mkv
------------------------------------------------------------------------------
VOB, 1 video track, 3 audio tracks, 3 subtitle tracks, 1:53:11
1: Joined VOB file
2: MPEG2, 480i60 /1.001 (16:9)
3: AC3, 2.0 channels, 192kbit/s, 48khz, dialnorm: -26dB
4: AC3, 2.0 channels, 192kbit/s, 48khz, dialnorm: -25dB
5: AC3, 2.0 channels, 192kbit/s, 48khz, dialnorm: -23dB
6: Subtitle
7: Subtitle
8: Subtitle
[v02] Extracting video track number 2...
[a04] Extracting audio track number 4...
[a05] Extracting audio track number 5...
[v02] Muxing video to Matroska...
[a03] Extracting audio track number 3...
[a05] Removing dialog normalization...
[a04] Removing dialog normalization...
[a03] Removing dialog normalization...
[v02] The MPEG2 stream changed from 60000i to 48000i at runtime 0:00:02.
[s07] Extracting subtitle track number 7...
[s08] Extracting subtitle track number 8...
[s06] Extracting subtitle track number 6...
[a03] Creating file "t - 3 - AC3, 2.0 channels, 192kbps, 48khz.ac3"...
[a04] Creating file "t - 4 - AC3, 2.0 channels, 192kbps, 48khz.ac3"...
[a05] Creating file "t - 5 - AC3, 2.0 channels, 192kbps, 48khz.ac3"...
[s08] Creating file "t - 5 - Subtitle.sup"...
[s06] Creating file "t - 3 - Subtitle.sup"...
[s07] Creating file "t - 4 - Subtitle.sup"...
Added fps value to MKV header.
Video track 2 contains 162910 frames.
eac3to processing took 1 minute, 50 seconds.
Done.

The "stream changed from 60000i to 48000i " is the consistancy. The mkv file produced is jerky during playback and this happens to all that I try to convert where this exact stream change occurs. Is there an option on eac3to that will solve this or is this a bug?

[jchappo]

Madashi,

Next time I get a Bluray in with LCPM audio, and a DTS-HD track, I will do more testing.

But for now, your assumption is incorrect. All the tracks I am converting are from full length feature films. The TrueHD, DTS-HD, and DD+ tracks come out roughly the same size, between 1 and 3gb depending on movie length. Once uncompressed to WAV, all files are > 4gb, even the TrueHD.

My question is, why do the LPCM tracks that ship with the Blurays work perfectly, while ones converted using eac3to don't? What is difference between the LPCM provided by the movie studio, and the one your program is making? Is there some kind of program I can use to analyze the two LPCM files?

Bascially, I wan't to do something similar listed in this thread:
http://forum.doom9.org/showthread.ph...533#post968533

But I want to go from DTS-HD/TrueHD/DD+ to LPCM. Would I use the RAW output option on eac3to, and map the channels like it says in that thread?

[nautilus7]

Originally Posted by jchappo

Madashi,

Next time I get a Bluray in with LCPM audio, and a DTS-HD track, I will do more testing.

But for now, your assumption is incorrect. All the tracks I am converting are from full length feature films. The TrueHD, DTS-HD, and DD+ tracks come out roughly the same size, between 1 and 3gb depending on movie length. Once uncompressed to WAV, all files are > 4gb, even the TrueHD.

My question is, why do the LPCM tracks that ship with the Blurays work perfectly, while ones converted using eac3to don't? What is difference between the LPCM provided by the movie studio, and the one your program is making? Is there some kind of program I can use to analyze the two LPCM files?

It seems that you didn't understand what madshi told you. Read again and tell what movie is that you 're converting. Maybe someone else can try it also.

[DeepBeepMeep]

Madshi,

When converting a TrueHD track to a DTS track with a multiparts seamless branching .m2ts file, are two passes really needed?

In my example I try to extract two tracks: an AC3 track and a TrueHD track (converted to DTS). Overlap is only detected on the AC3 track, is this normal? If so, could you automate the two passes to avoid calling surcode twice ? In fact, if the second pass was handled directly by EAC3TO you could save a lot of time since the demuxing has been already done.

Thanks for your great tool!

[tebasuna51]

Originally Posted by jchappo

My question is, why do the LPCM tracks that ship with the Blurays work perfectly, while ones converted using eac3to don't? What is difference between the LPCM provided by the movie studio, and the one your program is making? Is there some kind of program I can use to analyze the two LPCM files?

Like madshi say you the difference in only the header, the LPCM is raw data without header and the length is passed by other method.
The wav file >4GB have a field to inform about the length than must be wrong because have only 4 bytes and the max number is 2^32.

There are soft to manage these big wave files, you can encode to ac3 with Aften -readtoeof 1, or to mp4 with Neroaacenc -ignorelength. You can split, change bitdepth, samplerate, ... with WaveWizard. An also can be managed in AviSynth (BeHappy, SoundOut, Wavi) if you open the file with RaWavSource() method.

You can also convert (SoundOut) the header to w64 (Sonic Foundry, now acquired by Sony) format with fields than support until 2^64 filelength, but not all soft can accept this format.

[nautilus7]

Originally Posted by DeepBeepMeep

When converting a TrueHD track to a DTS track with a multiparts seamless branching .m2ts file, are two passes really needed?

In my example I try to extract two tracks: an AC3 track and a TrueHD track (converted to DTS). Overlap is only detected on the AC3 track, is this normal?

Yes, it is normal. Overlaps for TrueHD are really small and as a result not often met. If no overlaps are detected for a track, the 2nd pass isn't needed (there's no .gaps file created fo that track).

Originally Posted by DeepBeepMeep

If so, could you automate the two passes to avoid calling surcode twice?

Even if you needed 2 passes in your case, to avoid calling surcode twice (it is very slow), i suggest converting truehd (or whatever) to wav, do the 2nd pass (if needed) and then run a simple command to make the wav 2 dts encoding. This is what i always do when i want to encode to dts. Another good reason for feeding surcode direclty with wav files is to avoid the "surcode button didn't seem to work" error when you specify long filenames/paths as inputs. In that case you won't need to decode the source track twice.

[DeepBeepMeep]

Nautilus7,

Thans for your answer. I think it would be greatly appreciated anyway to automate the two passs processes since there is little reason why one would not go for the second pass when overlaps are discovered.

[tebasuna51]

Originally Posted by jchappo

Bascially, I wan't to do something similar listed in this thread:
http://forum.doom9.org/showthread.ph...533#post968533

But I want to go from DTS-HD/TrueHD/DD+ to LPCM. Would I use the RAW output option on eac3to, and map the channels like it says in that thread?

You don't need xport, sox, wavewizard or mediacoder, eac3to do all the job.
The 5.1 channels are automatically remapped from LPCM/DTS-HD/TrueHD/DD+
The RAW is the same than the wav if you delete the first 44 bytes.

[jchappo]

Let me fully explain the problem I have so everyone can understand it better.

The Popcorn hour A-100, is a network media player which can play uncompressed Bluray audio/video. It will accept a M2TS stream and passthrough the LPCM audio to my receiver. Some Bluray movies come with an LPCM audio stream already on them, some do not. For the ones that do not, I want to use eac3to to convert the TrueHD/DTS-HD/DD+ to LCPM so the A-100 can play it. But, I have only been successful with TrueHD so far, both DTS-HD and DD+ produce files that are garbage after the first 20-30% of the movie.

This is the problem. My question is, why can't eac3to produce LPCM files like that ones the movie studio includes on some discs? If I could produce those exact files, then that would be awesome and everything would just work. I am trying to figure out why eac3to is not producing these same files, and how to fix it.

[jchappo]

Would it help if I posted a sample of a movie studio created LPCM file, and one decoded from a DTS-HD track from eac3to which has garbage audio after 30%?

[nautilus7]

You keep saying your "problem", but you didn't listen to what madshi already said.

Do a simple test: Instead of making multi channel wav files make mono wavs. (eac3to input.thd/dtshd/eac3/etc output.wavs). Then check the center channel of all tracks. I bet every single one would be fine.

[wildchild22]

I have the same problem with lpcm as the above poster. What we really need is a program that can accept wavs not an interleaved wav file and mux that with video like scenarist. That way the audio track will be under 4 gigs. either that or maybe madshi could add an option to split the demuxed video and audio file every 30 mins. That way we maybe able to use tsmuxer and append the file and the sound may be correct.

[jchappo]

nautilis7,

I tried what you said and you are correct. If by dump luck, the TrueHD stream decoded into a working WAV, then why can't I edit the value in the WAV header to make it always work?

I'm just trying to get a working solution to producing working LPCM files.

[tebasuna51]

Originally Posted by wildchild22

I have the same problem with lpcm as the above poster. What we really need is a program that can accept wavs not an interleaved wav file and mux that with video like scenarist. That way the audio track will be under 4 gigs. either that or maybe madshi could add an option to split the demuxed video and audio file every 30 mins. That way we maybe able to use tsmuxer and append the file and the sound may be correct.

Then the problem is in tsmuxer, must accept wav > 4GB or multiple monowavs or w64/RF64 wav headers.

[EPiPH0NE]

Originally Posted by wildchild22

I have the same problem with lpcm as the above poster. What we really need is a program that can accept wavs not an interleaved wav file and mux that with video like scenarist. That way the audio track will be under 4 gigs. either that or maybe madshi could add an option to split the demuxed video and audio file every 30 mins. That way we maybe able to use tsmuxer and append the file and the sound may be correct.

I have already asked madshi about this and have been already trying to do this so I'll let you guys take over. Splitting the M2TS up then processing works for DTS-MA -> WAV but they can't be rejoined at this current time. We DO need to know how to make these files just like the studios are doing cause we are obviously missing something here. Maybe we should do:

eac3to.exe source.m2ts destination.wavs

then use another encoder to make the multi channel PCM/WAV? But, what encoder and what settings and how do we ensure proper LPCM channel configuration?

[Beastie Boy]

Have you tried

eac3to source.m2ts destination.raw

and then changing the file extention if necessary. Maybe this will get around any incompatabilities with the wav header.

Cheers, Beastie.

[madshi]

Originally Posted by gregt

One issue that I am having with DVDs is that the language seems to be missing for audio and subtitles

Just as a quick follow up:

The VOB format allows to have language identifiers inside the VOB stream. If there are any, eac3to will read them and display them. I do have some DVDs where language detection works. However, there are many DVDs where the language information is not contained in the VOB itself, but in the other files (*.IFO, I believe). eac3to currently doesn't parse these files, so the language cannot be displayed in such cases.

Originally Posted by sshd

The Machinist HD DVD:
3: h264/AVC, 1080p24 /1.001 (16:9) with pulldown flags

This title does not appear to have pulldown flags.
Video is destroyed when removing pulldown with h264info and plays fine at 23.976 fps without.

As far as I know, every single HD DVD movie (regardless of codec) does have pulldown flags. I think it's even a requirement of the HD DVD specification! If video is destroyed when using h264info then that probably indicates a bug in h264info. I'd suggest uploading a little sample and letting the h264info programmer know about the problem. He should be able to easily fix it.

Originally Posted by calinb

Indeed! Oftentimes, source that produces the "This track is not clean...Please clean the track with delaycut" error can be processed with eac3to ver. 1.16--after remuxing/joining the .evo(s) with EVOdemux, if necessary.

v1.16? That feels like ages ago...

Can you upload a little sample which works with v1.16 but not with v2.45?

Originally Posted by gregt

I seem to have run into a consistent problem with certain DVDs that are 480i60. [...] The "stream changed from 60000i to 48000i " is the consistancy. The mkv file produced is jerky during playback and this happens to all that I try to convert where this exact stream change occurs. Is there an option on eac3to that will solve this or is this a bug?

This is why I stated in the v2.45 release notes that I still have to further improve eac3to for MPEG2 clips with such mode changes. For now I recommend to either wait with converting such clips. Or to demux video instead of muxing it to MKV. I'm planning to complete the MPEG2 muxing improvements in the next build, which should then handle such mode changes just fine.

Originally Posted by DeepBeepMeep

When converting a TrueHD track to a DTS track with a multiparts seamless branching .m2ts file, are two passes really needed?

I cannot do it in one pass. The 2 passes are necessary to even out timecode instabilities.

Originally Posted by DeepBeepMeep

In my example I try to extract two tracks: an AC3 track and a TrueHD track (converted to DTS). Overlap is only detected on the AC3 track, is this normal?

Yes.

Originally Posted by DeepBeepMeep

If so, could you automate the two passes

I've already thought about this. Maybe I will.

Originally Posted by nautilus7

Do a simple test: Instead of making multi channel wav files make mono wavs. (eac3to input.thd/dtshd/eac3/etc output.wavs). Then check the center channel of all tracks.

Good idea...

Originally Posted by wildchild22

What we really need is a program that can accept wavs not an interleaved wav file and mux that with video like scenarist. That way the audio track will be under 4 gigs. either that or maybe madshi could add an option to split the demuxed video and audio file every 30 mins. That way we maybe able to use tsmuxer and append the file and the sound may be correct.

Originally Posted by jchappo

If by dump luck, the TrueHD stream decoded into a working WAV, then why can't I edit the value in the WAV header to make it always work?

I'm just trying to get a working solution to producing working LPCM files.

Originally Posted by tebasuna51

Then the problem is in tsmuxer, must accept wav > 4GB or multiple monowavs or w64/RF64 wav headers.

I agree with tebasuna51. This is clearly a problem with tsMuxer. So why should I spend hours and hours on adding fancy and complicated features to eac3to just to workaround a simple little "bug" in tsMuxer? Instead you guys should really pester the tsMuxer programmer about adding support for big WAV files and/or for multiple mono WAV files. That will solve all your problems at once and it will furthermore make things much easier for you, as well, cause you won't have to deal with manually splitting and rejoining files, anymore.