DGMPGDec 1.6.0 Development thread

[Guest]

Please enter your feature requests and bug reports for version 1.6.0 here. I would appreciate it if you have already entered them in the 1.5.0 final thread if you could repeat them here. Thank you.

[stax76]

if you could repeat them here. Thank you.

Making the CLI as robust that not the most wicked from mars filename can possibly break it, example for a valid filename on windows:

MPG[हिन्दी]-IF=,,[,,-IF=,,],,हिन्दी,,,хрватски.MPG.MPG

Dunno about C++ but with VB .NET it's possible to build a robust parser in a couple of minutes.

[burfadel]

An updated mpeg2 decoder. I believe the current one is based on an old decoder, and the decoding part is the same? I'm assuming there are many optimisations that can be done to it to make it noticeably faster for HD use (maybe even multithread it). I realise that this part is not done by yourself, but I'm sure there would be some keen people out there with programming knowledge that would be keen to update the decoder. Its not that the current decoder is slow, just that due to its threading etc. etc. it would have a noticeable impact on the final encoding speed.

[Tima]

A multithreaded indexer would be nice. Sometimes, the indexing speed is the bottleneck, not the speed of hard drive. E.g. when VOB is in windows file cache and you have no swap, so it's in RAM. :)

Another request is to write information in documentation about the offsets in Location value -- as for me, I don't understand what units are they in :(


Also repeating bug from 1.5.0 thread: http://forum.doom9.org/showthread.php?p=1146938#post1146938

[vlada]

I have two requests for command line:

1) Allow to display video output during indexing even if it is started from command line.

2) Add command line switch to index only a part of video.

Both things are possible from GUI so this shouldn't be a lot of work to implement (at least I hope so).

[=A=RGOS]

Hi,

For this thread, I propose a port of DGMPGDec to Linux :)
But, if you separe the program into a two part (GUI and command line) it's possible.

Bye,

[Sharc]

Would it be possible to include - as an option - auto-correction of fishy audio delay values (such as 87650ms for example)?

[GrofLuigi]

A "close" menu item.

GL

[Zep]

An updated mpeg2 decoder. I believe the current one is based on an old decoder, and the decoding part is the same? I'm assuming there are many optimisations that can be done to it to make it noticeably faster for HD use (maybe even multithread it). I realise that this part is not done by yourself, but I'm sure there would be some keen people out there with programming knowledge that would be keen to update the decoder. Its not that the current decoder is slow, just that due to its threading etc. etc. it would have a noticeable impact on the final encoding speed.

this is my #1 request also. multi threaded decoder would be great since for me the decoder is the bottle neck in my avs'.

[sunbeam]

Compatibility with W2K as continued in 1.5.x would be great :)

[fbgd]

Hi,

For this thread, I propose a port of DGMPGDec to Linux :)
But, if you separe the program into a two part (GUI and command line) it's possible.

Bye,
I've never had a problem running it with wine. Plus wouldn't that require a native linux version of avisynth?

[Guest]

Adding this request in PM from morphinapg:

Thank you and I'm happy to tell you I've been using the

__aud__

to get the first audio track to use for my ASXGui and it works great. But someone just requested a feature to include support for the tracks other than the first. So my question is, keeping with the same idea that the track order is decided by the track ID, could it be possible to add things such as

__aud2__
__aud3__
__aud4__

etc up to maybe 8 tracks or so, and if there weren't that many tracks then they would simply not return anything.

It's not a huge priority just maybe something I would like to see down the road as it would be helpful for my program, and other people may benefit from it as well.

[=A=RGOS]

I've never had a problem running it with wine. Plus wouldn't that require a native linux version of avisynth?

Yes it work fine on wine :), but a native avisynth version is under developpement (v3.0).
And a DGMPGDec on Linux with native build is more appropiate for certain use.

bye bye

[hajj_3]

just wondering if something changed in final build of dgmpgdec, when i open a .mpg and choose save project normally it creates a .mpa file as i changed the option in dgindex.ini from .mp2 but it doesnt create any by default, i have to choose "demux all tracks" instead of the default "demux track". I dont remember having to change this in any previous builds, did this change in the final build?

everything is running nicely with this build btw, has done since RC5.

[Guest]

The Select Tracks dialog defaults to empty. If you choose Demux Tracks, then you have to specify one or more tracks in the Select Tracks dialog.

[smok3]

a. if i understand correctly, wav files decoded from ac3 are allready delay corrected? Same request/wish for mp2 audio (decoder + delay correction if neccesary)

(I'am using it from cli as well)

b, another thing, basic info about the stream processed (video type, resolution, ipd, audio type, #of channels, # of tracks) dump to __info__ or similar var to be used with avisynth template. So my template could be, example;
a=mpeg2source("__vid__")
b=wavsource("__aud__")
audiodub(a,b)

__END__

__info__

[guada2]

Hello neuron2,

Just 2 things: :)
- MLP (audio)
- VC-1 (video)

[Underground78]

Hello,

This is not a serious bug but sometimes when I navigate in a TS file using the "<" button to go the previous I-Frame, I get stuck at a frame and I can't go back in the file until I use the navigation bar. If I go to the next I-Frame using the ">" button when I'm stuck, it works but if I retry to go back I get stuck at the same frame. When I have such a frame in a file, the bug can always be reproduced. I've never seen this bug when I navigate next I-Frame to next I-Frame.

I've tried to give you a sample but when I demux the video with DGIndex, I can't reproduce the bug in the resulting file. If there is another way to cut the file, I would be happy to try ! :)

[Guest]

There have always been issues with stepping backwards by GOP. I'm aware of them but it's cosmetic so likely I won't do anything about it as it won't displace higher priority things. It's hard to give me a stream because if you cut from the beginning it changes things such that the problem may not be triggered.

[Underground78]

There have always been issues with stepping backwards by GOP. I'm aware of them but it's cosmetic so likely I won't do anything about it as it won't displace higher priority things. It's hard to give me a stream because if you cut from the beginning it changes things such that the problem may not be triggered.

Ok, no problem ! :)

And just to know what are your plans for DGMPGDec 1.6.0 ?

[Guest]

No plans yet. That's what this thread is for.

[Sharc]

Don't know if this is a bug or feature of DGIndex.
I have a VTS with 2 PGCs: PGC1=Movie, PGC2=studio intro (separate VOB ID with 1 cell of 12 seconds duration).

Demuxing with DGIndex produces 2 audio tracks: 0x80(en) and 0x81(de).
The demuxed (0x80 en) stream includes the (en) audio of the movie and the (en) audio of the studio intro. DELAY= 0ms.
The demuxed (0x81 de) stream includes the (de) audio of the movie only, but no audio for the studio intro. DELAY =-16ms.
Apparently the studio logo uses the same audio stream 0x80 for both languages (en and de), as reported by VobBlanker for example.

The problem comes with re-muxing: The (de) audio is then 12 seconds out of sync (ahead of the video) because of the missing audio of the studio intro in the 0x81 stream, or due to the low delay indication of -16 ms instead of +12 seconds. Don't know if and how DGIndex can or should handle such cases.

Added:
Workaround id to step with > button to the first cell of the movie and pressing [, then Save Project, i.e. skipping the intro pgc/vob id. This will produce synchronous audio tracks with same DELAY for the movie.

[Malow]

perfect sync with broken HDV clips? :p

[Sharc]

Wish for future release:
- Option for indexing and demultiplexing split according to PGCs, not only on a per file basis (including all PGCs)
- Angle selection

[morphinapg]

I have some HD recordings that switch between 5.1 AC3 and 2.0 AC3 between commercials. Is it possible to have an option to demux all of the 2.0 AC3 to one file and all of the 5.1 AC3 to another (with empty space filling the rest)? Either that or split on audio format change. I'd prefer the first option, but I could live with the second.

Otherwise I have horrible sync issues that are often not fixable.

[Guest]

Adding request in PM from LordIntruder:

Some months ago I ripped a movie, PAL and progressive, around 1h50 long. From 1h40 to 1h45 suddenly the sequence was interlaced and then the last 5 minutes progressive like the rest of the movie.

This week I encountered the exact same problem with an old serie. In the middle of the episode, around 10 min, suddenly everything is interlaced.

Could you implement such an option to tell us automatically that the video has both progressive and interlaced content? Because as usual, I always see that AFTER the encoding Grrr...

I thought that maybe when we save the project (F5), at the end in the log file, we could have a report saying:

- Both progressive and interlaced

Or even greater:

Say a movie has 150000 frames.

- frame 1 to 57999: progressive
- frame 58000 to 65000: interlaced
- frame 65001 to end: progressive

Here we could see immediately where in the stream the interlaced content is, go to that area and test a small encode with the best deinterlace settings.
------------


So I don't know if this can be done or not, I have no knowledge about that but if this is possible, it would be very great.

[Ranguvar]

Definitely what LordIntruder suggested... also, please more info (with frame/time ranges) of what's flagged as telecined, etc.

[Zep]

Definitely what LordIntruder suggested... also, please more info (with frame/time ranges) of what's flagged as telecined, etc.

yes yes I third that! that would be a very awesome feature since you could use trim and or range to process those areas differently and with hybrid source you could target areas with a slower but much better frame blender/double.

[stax76]

@neuron2

The -minimize cmdl switch isn't really necessary I think. The window state can be defined launching a external process, even from the cmdl shell using the START command. Currently minimizing does not work launching DGIndex from code as only the main window minimizes but not the preview dialog, minimizing manually the preview dialog hides, minimizing any other method it don't hide. I was thinking maybe there is a way I can start DGIndex from code without the preview to show, first thing I tried starting indexing manually and clicked the preview away, this causes bug #2, DGIndex is freezing.

[LoRd_MuldeR]

I got a question: When I load an MEPG PS file (e.g. captured from my DVB-S card), save the project and demux the audio, I get an audio file like "FOOBAR T01 DELAY XXXms.mpa". This delay value is very useful to me, because it helps to synchronize the audio. But I don't need the audio file, nor do I need the D2V file. I do all my work on the MPEG file. All I need is the delay value. So is there a way to simply display that delay value in DGIndex? If not, which I think is the case, could it be added please? Like "Tools > Calculate Audio Delay" or something...

[Sharc]

I wish under "Options" a 16 x 16 grid overlay be shown. It would help for cropping and estimating how the useful picture fits the 16 x 16 macroblocks.

[Revgen]

Can you add a BFF/TFF switch?

I cap HDTV sports programs and the field order keeps changing from TFF to BFF. The D2V repair tool works fine to fix that, but it fixes it according to the field order at the beginning of the video. I always want my resulting .d2v file to be TFF. I currently can do that using Stickboy's JDL_ReverseFieldDominance(), but it would be nice if it could be changed within DGIndex.

Thanks.

[Ranguvar]

And, could you add a line to the log file that reports where (in frames) any corrected field order transitions occurred?

[GrofLuigi]

Support for Sharc's idea - 16X16 grid (switchable, of course) and Revgen's idea (force the output stream to become TFF/BFF only). :thanks:

GL

[mikeytown2]

I'll go ahead and request this for SynchronousArts and spyros78. Reading of the SMPTE timecodes (http://forum.doom9.org/showthread.php?t=138139).

Me personally, I also would like the decoder to be multi threaded.

[krosswindz]

I just upgraded to DGIndex 1.5.0 I am just wondering if this was a problem with DGIndex 1.4.9 or is it a bug in DGIndex 1.5.0

The same DVDs colorimetry information is different. 1.4.9 reports bt709 while 1.5.0 reports the sames bt407

v1.5.0
http://img66.imageshack.us/img66/8110/dgi15zg2.jpg

v1.4.9
http://img354.imageshack.us/img354/723/dgi149bg8.jpg

[Guest]

Did you read the change log? Change number 21 would be germane.

http://neuron2.net/dgmpgdec/changes150.html

[krosswindz]

Did you read the change log? Change number 21 would be germane.

http://neuron2.net/dgmpgdec/changes150.html

makes sense now... :thanks:

[Ranguvar]

The ability to demux the PCM WAV audio from Blu-ray M2TS files... :)

[SynchronousArts]

I'll go ahead and request this for SynchronousArts and spyros78. Reading of the SMPTE timecodes (http://forum.doom9.org/showthread.php?t=138139).

@mikeytown2,

Thanks for adding this here. I added some idea's to that thread about how the Timecode could function.
I recently was given a project with about 20 HDV camera tapes that have discontinuous Timecode all over them.
Would have been a big help to have this working.

Thanks

SA

[Sharc]

Would it be possible to add in the information - and log file - the Pixel Aspect Ratio (PAR, SAR) of the source? If it is not specified in the stream one could make a best assumption marked with an * ( similar to the colorimetry).

[Blue_MiSfit]

Hey folks,

Wondering if DGIndex supports demuxing SMPTE 302M (AES/EBU audio stored in MPEG-2 Transport Stream)?

I may have to handle some of this stuff in the future, and it would be lovely if DGIndex could automagically handle it ;)

~MiSfit

[Guest]

Do you have the specification?

[mikeytown2]

looks like someone attempted demuxing SMPTE 302M back in 2006 for ffmpeg
http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2006-September/thread.html#15650

Dead end over on hydrogenaudio, only 3 posts
http://www.hydrogenaudio.org/forums/index.php?act=Search&CODE=show&searchid=f7b75b51c414102a2e1e5b82e19d4db8&search_in=posts&result_type=posts&highlite=SMPTE+302M

The specs looks like it will cost $35
http://store.smpte.org/SearchResults.asp?Search=302

or $22 for a 2002 edition
http://www.techstreet.com/cgi-bin/detail?product_id=1249916

Or wikipedia
http://en.wikipedia.org/wiki/AES/EBU

Which lead to page 22 of this pdf
http://www.ebu.ch/CMSimages/en/tec_AES-EBU_eg_tcm6-11890.pdf
this might be enough to get an implementation going...?

Library Of Congress is of no help
http://www.digitalpreservation.gov/formats/fdd/fdd000208.shtml

[Blue_MiSfit]

Wow - thanks for doing research for me mikeytown2!

I don't have the spec unfortunately... if I did I would certainly forward it

~MiSfit

[Guest]

You can buy it if it is that important to you.

[Blue_MiSfit]

Indeed I will if the need arises!

[CruNcher]

Donald http://libmpeg2.sourceforge.net/ it was recently updated twice do you think it makes sense to update the core of Dgdecode now with it or are the changes not so important ?

[Guest]

DGDecode uses an internal decoder, not libmpeg2. It would be a major rewrite to replace it.

[stattik]

Hi neuron2,

How about support for decoding AC3 to WAV for TS files?