eac3to - audio conversion tool

[madshi]

Originally Posted by jchappo

I will do more testing, but so far, the only tracks I can play on the popcorn hour are:

LPCM tracks included with the Bluray
TrueHD converted to WAV (PCM) with eac3to -libav

Both DTS-MA and DD+ (Sonic, and Nero) both produce bad WAV files.

That is not a conclusive test. The bad WAV playback probably only depends on the WAV file size and on nothing else. Most probably the WAV files created from DTS-MA and DD+ are bigger than 4GB while the TrueHD created WAV file is probably smaller than 4GB. Probably that's the difference...

[madshi]

eac3to v2.45 released

http://madshi.net/eac3to.zip

* Blu-Ray angles are now reported as separate titles
* duplicate playlists are not listed in the "folder view", anymore
* reduced TrueHD and RAW/PCM gap/overlap threshold to 7ms
* reduced (E-)AC3 gap/overlap threshold to 60% of the runtime of one audio frame
* reduced MP2 gap/overlap threshold to 60% of the runtime of one MP2 frame
* reduced DTS threshold to 60% of the runtime of one DTS frame, but at least 7ms
* fixed: Blu-Ray chapter export sometimes wrote incorrect "00:00:00.000" items
* improved handling of MPEG2 streams (changes from interlaced to progressive)
* video information now shows "with pulldown flags", if applicable
* removed "-ignoreDiscon" from help; hint is shown when a discontinuity occurs
* added "-ignoreEncrypt" option; hint is shown when a source is encrypted
* new option "-extensible" creates WAV files with a slightly different header
* fixed some smaller bugs

Notes:

(1) As demanded, I've noticably decreased the audio gap/overlap thresholds. For most audio tracks the threshold is only 7ms now. If you run into any trouble with this (e.g. gaps/overlaps reported in situations where you didn't expect it), please let me know. Also if it works well for you, I'd also like to hear about it and maybe see some success logs. Thanks!

(2) Please note that MPEG2 handling of problematic streams has improved, but I still not to improve it further. As an explanation: Many MPEG2 streams are 60i. Some of them are natively interlaced, others are movies. Movies are often encoded as 24p but with pulldown flags. HD DVD uses this method. Now the problem is this: Sometimes a stream begins as 24p with pulldown flags but then suddenly switches to natively 60i interlaced. Or the other way round. Some streams even switch multiple times. Of course we want eac3to to handle 24p movies with pulldown flags as 24p. But doing this runs into problems when the stream suddenly switches to 60i interlaced content. Currently any stream with changes between 24p (with pulldown flags) and 60i is not handled in the best possible way by eac3to. I still need to improve that. But at least processing is not aborted, anymore, if there's such a change in the stream. The previous eac3to version aborted in such a situation. That means with v2.45 you can at least use eac3to for demuxing video and for handling the audio streams etc...

(3) I've heard that TsMuxer complains about eac3to's WAV files. Something like "channel mapping information missing". You can now use the new "-extensible" switch to make eac3to write slightly different WAV headers. This may satisfy TsMuxer (or not).

[jchappo]

Actually the TrueHD WAV file was 6GB, DD+ and DTS-HD both produce WAV files around that size as well.

[markrb]

[a04] Encoding AC3...
[a04] Creating file "D:\audio.ac3"...
[a04] Audio overlaps for 14ms at playtime 0:00:02.
[a04] Audio overlaps for 25ms at playtime 0:05:09.
[a04] Audio overlaps for 12ms at playtime 0:09:18.
[a04] Audio overlaps for 5ms at playtime 0:09:53.
[a04] Audio overlaps for 29ms at playtime 0:12:14.
[a04] Audio overlaps for 23ms at playtime 0:13:32.
[a04] Audio overlaps for 10ms at playtime 0:26:09.
[a04] Audio overlaps for 23ms at playtime 0:28:19.
[a04] Audio overlaps for 32ms at playtime 0:34:52.
[a04] Audio overlaps for 24ms at playtime 1:03:31.
[a04] Audio overlaps for 8ms at playtime 1:04:06.
[a04] Audio overlaps for 11ms at playtime 1:20:23.
[a04] Audio overlaps for 7ms at playtime 1:22:20.
[a04] Audio overlaps for 13ms at playtime 1:24:06.
[a04] Audio overlaps for 14ms at playtime 1:26:21.
[a04] Audio overlaps for 21ms at playtime 1:29:26.
[a04] Audio overlaps for 11ms at playtime 1:29:54.
[a04] Audio overlaps for 21ms at playtime 1:37:21.
[a04] Audio overlaps for 28ms at playtime 1:39:05.
[a04] Audio overlaps for 27ms at playtime 1:39:27.
[a04] The audio file was demuxed without making use of the gap/overlap information.
[a04] Please rerun the same eac3to command line. That will correct the gaps/overlaps.

I don't understand what it's asking me to do.
Am I suppose to just run the program again and it will somehow remember and fix the issue or am I suppose to put in some sort of command to fix it?

Thanks,
Mark

[madshi]

Originally Posted by jchappo

Actually the TrueHD WAV file was 6GB, DD+ and DTS-HD both produce WAV files around that size as well.

Ok. Then please try decoding the very same audio track with all 3 decoders and compare. I'd be surprised if it depended on the decoder which WAV plays fine for you and which doesn't. But of course anything is possible.

[Yraen]

Originally Posted by markrb

[a04] The audio file was demuxed without making use of the gap/overlap information.
[a04] Please rerun the same eac3to command line. That will correct the gaps/overlaps.

I don't understand what it's asking me to do.
Am I suppose to just run the program again and it will somehow remember and fix the issue or am I suppose to put in some sort of command to fix it?

Thanks,
Mark

Yes, just run the same command line again.

[madshi]

Originally Posted by markrb

I don't understand what it's asking me to do.
Am I suppose to just run the program again and it will somehow remember and fix the issue

Yes. You're supposed to simply rerun eac3to with exactly the same parameters as in your first try. The information needed for fixing the issue were stored on harddisk during the first pass (look for a *.gap file). This is then made use of when you run eac3to a 2nd time.

If you want you can save the original "d:\audio.ac3" file and compare it with the final result. You should notice that the current "d:\audio.ac3" file is not in sync with the video. The AC3 file you get from the 2nd pass should be in sync.

Edit: Yraen was faster...

[bmnot]

"The h264 muxer received invalid h264/AVC data" error problem fixed as expected. Thanks.

EDIT: I have a potentially silly question. If I have a 16-bit lossless source I want to convert to DTS, should I encode the DTS at 16-bit, because there is no need for 24, and so the size will be smaller? Or would encoding to DTS at 24-bit help the sound quality?

[nautilus7]

Surcode will encode at 24 bit. You don't have/need to do anything.

[Thunderbolt8]

thanks for the new version

2 questions:

"reduced DTS threshold to 60% of the runtime of one DTS frame, but at least 7ms" - does this affect dts-hd (MA) as well?

and regarding those mpeg2 sources, which sometimes switch flags or show other problems, should I still continue to make samples of these? or do you have already enough and just need more time for those improvements atm?

[Atak_Snajpera]

Is it possible to demux only chapters?

[Bluestraw]

Many thanks for the new version - I tested the angle support, and successfully converted the French version of Cars (which is at least 12 parts) to MKV with FLAC, and the resulting file is perfectly in sync. Only reported a couple of small overlaps, each 4 ms I think.

[Yraen]

Originally Posted by Atak_Snajpera

Is it possible to demux only chapters?

Yes. Most chapters are track 1, but check to be sure.

eac3to H:\moviefolder\ 1) 1: chapters.txt

[jchappo]

Ok Madashi,

Converted the DTS-HD track to WAV with all decoders, all suffered from same problem, after 20-30% it was all garbage.

Same thing converting from DD+.

Only thing that works is TrueHD.

[nautilus7]

Originally Posted by jchappo

Ok Madashi,

Converted the DTS-HD track to WAV with all decoders, all suffered from same problem, after 20-30% it was all garbage.

Same thing converting from DD+.

Only thing that works is TrueHD.

There must be something wrong with your installation of the sonic decoder. Which version do you have? Which movie did you try?

[gregt]

Thanks for this update (2.45)!
- Fixed the Xmen3 problem which now converts successfully.
- Fixed the DVD interlaced problem and my interlaced (480i60 and 480p24 with pulldown flags) files now process with no problems (so far)! Most of my DVDs the changed occurs in the first 2 seconds.
- the gap change seems to create a lot gap data where there was none before. I ignore it on files that I know it is not needed.
One issue that I am having with DVDs is that the language seems to be missing for audio and subtitles as follows:
eac3to v2.45
command line: eac3to vts_01_1.vob+vts_01_2.vob+vts_01_3.vob+vts_01_4.vob+vts_01_5.vob t.mkv
------------------------------------------------------------------------------
VOB, 1 video track, 2 audio tracks, 3 subtitle tracks, 13:17:28
1: Joined VOB file
2: MPEG2, 480p24 /1.001 (4:3) with pulldown flags
3: AC3, 2.0 channels, 192kbit/s, 48khz, dialnorm: -27dB, -83ms
4: AC3, 2.0 channels, 192kbit/s, 48khz, dialnorm: -27dB, -100ms
5: Subtitle
6: Subtitle
7: Subtitle

But, thanks for the update as that is a huge improvement for me.

[nautilus7]

To make eac3to list languages you have to load the DVD folder, like you do for HD DVDs and Blu-ray Discs. Reading DVD folders isn't supported (yet), though, so you can't do it currently.

[PHD_1976]

Originally Posted by madshi

Maybe (if it's not too much work) it would make sense to add this functionality to eac3to? Or are there more reasons why you want to create your own tool? Not that it would be a problem for me, of course!

Madshi,
There's a lot of work fixing h264 streams for Scenarist.
Maybe we can discuss it outside this thread.
I don't mind sharing my code with you.
Where can I write you a letter with a description of what should be done to fix H264 streams?

[madshi]

Originally Posted by Thunderbolt8

"reduced DTS threshold to 60% of the runtime of one DTS frame, but at least 7ms" - does this affect dts-hd (MA) as well?

Yes.

Originally Posted by Thunderbolt8

and regarding those mpeg2 sources, which sometimes switch flags or show other problems, should I still continue to make samples of these? or do you have already enough and just need more time for those improvements atm?

Right now I don't need further samples. Once I'm done with the improvements and there are still problems, I'd be interested in further samples. But for now I think I have enough samples.

Originally Posted by Bluestraw

Many thanks for the new version - I tested the angle support, and successfully converted the French version of Cars (which is at least 12 parts) to MKV with FLAC, and the resulting file is perfectly in sync. Only reported a couple of small overlaps, each 4 ms I think.

Good to hear, thanks!

Originally Posted by jchappo

Converted the DTS-HD track to WAV with all decoders, all suffered from same problem, after 20-30% it was all garbage.

Same thing converting from DD+.

Only thing that works is TrueHD.

Do you have different movies with DTS-HD, DD+ and TrueHD tracks? If possible, it would be helpful if you could test other movies, too.

I still believe that the size of the WAV file is the only problem here. Let me explain: The WAV header has a "size" field which is only 4 byte long. The max size you can put in 4 bytes is about 2GB or 4GB (depending on whether you interpret that number to be able to hold negative numbers, too). Now if you have a WAV file which is longer than 2GB or 4GB, the "size" field in the WAV header is invalid. It's difficult to explain to a non-programmer what value this field has when it's invalid. But you can think of it as being "random", but the random value depends on the full WAV file size. Now if a media player is stupid enough to rely on the "size" field, it will interpret the random value in the WAV header as being the correct size. Since the value is random, a 5.9GB WAV file could eventually work much better than a 6.1GB WAV file or vice versa.

When you decode the same track with different decoders, they will likely still output the same WAV size. That's why all 3 decoders behave the same for you. I believe the only problem is the random WAV "size" field. You probably just had luck with your TrueHD track that the random size field somehow made it work more or less ok.

Originally Posted by gregt

Fixed the DVD interlaced problem and my interlaced (480i60 and 480p24 with pulldown flags) files now process with no problems (so far)! Most of my DVDs the changed occurs in the first 2 seconds.
- the gap change seems to create a lot gap data where there was none before.

Yeah, that's still something I'm working on. It happens especially when a movie begins as 480p24 (with pulldown flags) and then changes to 480i60.

Originally Posted by gregt

One issue that I am having with DVDs is that the language seems to be missing

As nautilus7 already explained, this is a feature which is just not implemented in eac3to. eac3to parses the special HD DVD and Blu-Ray files where the language information is contained. But eac3to doesn't parse the special DVD files where the info is contained.

Originally Posted by PHD_1976

There's a lot of work fixing h264 streams for Scenarist.
Maybe we can discuss it outside this thread.
I don't mind sharing my code with you.
Where can I write you a letter with a description of what should be done to fix H264 streams?

You can email to dear (at) madshi (dot) net.

[PHD_1976]

Originally Posted by madshi

You can email to dear (at) madshi (dot) net.

Just emailed you from genetr (at) mail (dot) ru