DG NV tools

[CruNcher]

@Neuron2
Does nvcuvid.dll (VP2) support Mpeg-2 4:2:2 Studio Broadcast streams ?

[Guest]

Currently no, but funny you should ask. I am working with Nvidia on that right now.

[aegisofrime]

neuron2, I'm thinking of donating for your DG NV Tools and I have some questions I hope you can help me with.

The Purevideo deinterlacing function looks great, and that's what I guess I will be using most. The thing is my work computer is also my gaming computer. I'm currently using a Radeon 4870 and need an nVidia card. I have come up with 2 choices

1) Get a Fermi (GTX 470 is my first choice)

2) Continue using my Radeon 4870 and get a cheap GT210/GT220 and use it alongside with the 4870 purely for Purevideo.

My question is, is the purevideo deinterlacing quality dependent on the card or are there differences? Also, do you know if your software will work in scenario 2? I'm aware that only Windows 7 supports heterogeneous video card support. I'm currently on vista and will have to upgrade to 7 for that, but I hope to find out whether it will work first.

Thanks!

[squid_80]

I've been running ATI and nVidia cards in the same machine since windows XP, no need for windows 7.

[LigH]

Additional SH in GOP ... wild guess: To support decoders catching a new video stream after DVB channel switching?

[Guest]

Originally Posted by aegisofrime

My question is, is the purevideo deinterlacing quality dependent on the card or are there differences?

I don't know the answer to that. You'd have to ask Nvidia. I do know that it will be fine with the 210/220.

Also, do you know if your software will work in scenario 2? I'm aware that only Windows 7 supports heterogeneous video card support. I'm currently on vista and will have to upgrade to 7 for that, but I hope to find out whether it will work first.

Yes, that scenario is supported.

[Guest]

Originally Posted by LigH

Additional SH in GOP ... wild guess: To support decoders catching a new video stream after DVB channel switching?

No, because an SH doesn't make a P frame into an I frame, so that can't improve seeking at that point. Anyway the GOP is very small to begin with.

[Guest]

* The previously hidden transport stream trim functionality is now exposed in
the GUI. A progress dialog for the trim operation has been added. Performance of the
file write operation has been improved. The functionality is now more robust for
writing proper packet boundaries. 192 byte transport packet streams are now
properly supported. Refer to the users manual for important usage notes and
caveats.

* Added a CLI avs template command: -at file.

* Fixed the Disable Display function for double rate deinterlacing. The operation
of this function has been made more robust.

* Fixed a problem causing some AAC streams to not be detected.

* Fixed a problem causing some teletext streams to be detected as audio.

* Added an error popup for non-4:2:0 streams (until supported by Nvidia).

* Fixed a problem in indexing of MPEG2 streams that caused some unusual but legal
GOPs to be considered malformed and cause an abort of the indexing.

http://neuron2.net/dgdecnv/dgdecnv.html

[Guest]

BTW, my 64-bit system is up and running now. Clean install and activate on bare metal with only a Win7 upgrade disk is quite a challenge, but I was up to it.

Development of 64-bit DGDecNV is begun.

[Clumpco]

Originally Posted by neuron2

BTW, my 64-bit system is up and running now... Development of 64-bit DGDecNV is begun.

Yay! ... and :sad: I can't afford an I7 and Win7 64 bits for the moment

[aegisofrime]

Originally Posted by neuron2

I don't know the answer to that. You'd have to ask Nvidia. I do know that it will be fine with the 210/220.

Yes, that scenario is supported.

Thanks for the reply If I do get a Fermi I will report back on its quality.

Originally Posted by squid_80

I've been running ATI and nVidia cards in the same machine since windows XP, no need for windows 7.

Thanks. I'm aware that heterogeneous is supported on XP as well, but I greatly prefer 7 to XP.

[Clumpco]

Version 2010 now successfully avoids treating Teletext as audio :-)
Unfortunately it still seems to detect BBC HD ac3 2/0 as 3/2. But no biggie.

[Guest]

Originally Posted by Clumpco

Unfortunately it still seems to detect BBC HD ac3 2/0 as 3/2.

Did you give me a stream for that? If so, what was the name? If not, please do so.

[djloewen]

I've got a video that isn't behaving optimally. It's an extra from a blu-ray disc, encoded at 720x480 (and it should be 4:3 but it's apparently not flagged as such, it appears to even display on a television in a slightly stretched AR). eac3to detected it as "480p24/1.001 (3:2) with pulldown flags". When it's indexed using DGindexNV 2010 it reports "Frame Rate: 23.976024 fps" but it seems to be counting 30fps worth of frames. As a result, when I encode the video I end up with ~22mins of audio and ~28mins of video (30fps played back at 24fps). Small sample here:
http://www.megaupload.com/?d=PW53CQ2O

Any help would be appreciated.

[Mixer73]

Originally Posted by neuron2

Development of 64-bit DGDecNV is begun.

64bit Vista here for testing of any builds you wish, although I am only using DGDecNV in a very limited way.

[Guest]

Originally Posted by djloewen

eac3to detected it as "480p24/1.001 (3:2) with pulldown flags".

I don't see any soft pulldown in the stream you linked.

When it's indexed using DGindexNV 2010 it reports "Frame Rate: 23.976024 fps" but it seems to be counting 30fps worth of frames.

The stream is indeed coded as 23.976 fps. Regarding your second point, I cannot tell anything without the muxed stream to compare the audio. It sounds unlikely.

Please post a link to a larger M2TS source sample direct from the bluray that I can use to duplicate the problem.

[Clumpco]

Originally Posted by neuron2

Did you give me a stream for that? If so, what was the name? If not, please do so.

Not recently, no. However I suspect that the .ts that it did it on was corrupt, after "fixing" DGNV detected it correctly as 2/0.

I shell get back to you after more investigation.

[djloewen]

Originally Posted by neuron2

I don't see any soft pulldown in the stream you linked.

The stream is indeed coded as 23.976 fps. Regarding your second point, I cannot tell anything without the muxed stream to compare the audio. It sounds unlikely.

Please post a link to a larger M2TS source sample direct from the bluray that I can use to duplicate the problem.

Here's a 100mb sample of the original m2ts file:
http://www.megaupload.com/?d=MKOFDA0W

Here's the full m2ts file, in case it helps (763mb):
[link removed]

This particular episode (there are 4) is 20:17, but the video ends up being 25:21.

And here's something interesting. Last night I was using CUVIDserver 2006 and DGindexNV 2010 (accidentally, due to MeGUI not using the newest version) and it said 23.976. This morning I used 2010 and 2010, and it said 29.970. So I've spent 15 minutes switching between various combinations (and switching between input files: the 100mb sample, the full m2ts, and the demuxed vc1) and in fact I'm not seeing a pattern at all - sometimes I get 23.976, sometimes 29.970. I've managed to get a DGI file of each, and the one created at 29.970 plays back at the correct length in an AVS file.

So, besides the obvious "what's going on here?", there's the interesting point that in either case, it counts the same number of coded frames and playback frames. Is this correct? Is eac3to wrong about it having soft pulldown?


Edit:
Alright, I've found the pattern (and btw I'm only using 2010 now, so I don't know how this applies to 2006). If I open one of the m2ts files first, it shows 29.970 and any of the test files I subsequently open (inclding the demuxed vc1 stream) show as 29.970. If I then close and re-open DGindexNV, and open a vc1 first, it detects as 23.976 and opening the m2ts afterwards will make it show as 23.976.

[Guest]

Originally Posted by djloewen

Alright, I've found the pattern (and btw I'm only using 2010 now, so I don't know how this applies to 2006). If I open one of the m2ts files first, it shows 29.970 and any of the test files I subsequently open (inclding the demuxed vc1 stream) show as 29.970. If I then close and re-open DGindexNV, and open a vc1 first, it detects as 23.976 and opening the m2ts afterwards will make it show as 23.976.

Thank you for your trouble report. I wish they were all as thorough and well presented.

I've duplicated this with your samples. Investigating...

[Guest]

The stream is 29.97 fps without pulldown.

Regarding the initial sample.vc1 you linked: I don't know what you demuxed that with but it really is 23.976. Let's forget that because I don't know how you made it. If EAC3TO did that you should report it. If you demux with DGIndexNV 2010 from your sample.m2ts, the resulting sample.demuxed.vc1 opens as 29.97, as it should.

I've located the bug that you exposed. When you open a new file from within DGIndexNV (without closing and reopening DGIndexNV), there are cases where the new sequence headers are not processed. One case is if the only change is the frame rate, which applies in your scenario because you open sample.vc1 first and then open sample.m2ts.

The fix will be in the next release. Your workaround is to close DGIndexNV before opening new files.

(BTW, I have deleted your full sample link for copyright reasons.)