DGVC1DecNV Version 1.0.6

[Guest]

DGVC1IndexNV today decoded its first VC1 stream from the blu-ray "Baraka". I'm still some time away from being able to release an alpha, but I wanted to start this thread to let you know about it and to ask for some test clips.

I'm looking for any and all VC1 clips, but only raw elementary VC1 or in an M2TS container. I will do ASF later. Thank you.

For those not acquainted with my tools, the NV on the end denotes a version that runs on an Nvidia graphics card GPU. I don't have any plans to develop a non-NV version because I am not aware of any decent open source decoder.

Also please be aware that my NV tools require a license obtained by making a small donation to become a supporter of neuron2.net.

EDIT: This post will now hold a link to the latest version.

http://neuron2.net/dgvc1decnv/dgvc1decnv.html

[Snowknight26]

Had a short clip laying around: http://stfcc.org/misc/jessejames.clip.m2ts

[Guest]

Thanks, Snowknight26. It plays just fine.

[lucassp]

What GPU did you use for decoding?

[Guest]

I used both 8500GT and 9600GT.

[Audionut]

00014.split.54.ts (http://www.greentreefrog.net.au/~lionel.davey/00014.split.54.ts)

I need to go through my blu-rays and see whats vc1.

[Audionut]

00403.m2ts (http://www.greentreefrog.net.au/~lionel.davey/00403.m2ts)

[Inventive Software]

I'm keeping a close eye on this one, as it'll help me justify whether I purchase a new graphics card for Xmas.

[rack04]

sample.m2ts (http://www.megaupload.com/?d=Z3FL0VMY)

Hopefully this works. Here is the procedure I followed to create this file.

1.) Source is evo container from HD DVD
2.) Extracted VC1 and E-AC3 using eac3to
3.) Muxed VC1 and E-AC3 to M2TS using tsMuxer

[lucassp]

I used both 8500GT and 9600GT.

Officialy those two cards don't support VC-1 full bitstream decoding. Or maybe nVidia updated the VP2 core:confused:

[Guest]

Huh? What do you mean by "full bitstream decoding"? Do you mean it's not all on the GPU? So what? We need an Avisynth frame accurate VC1 source filter, that's all, don't we? I get 43fps on 1080P VC1. Who cares how the decoding is partitioned?

[Guest]

sample.m2ts (http://www.megaupload.com/?d=Z3FL0VMY)

Hopefully this works. Here is the procedure I followed to create this file.

1.) Source is evo container from HD DVD
2.) Extracted VC1 and E-AC3 using eac3to
3.) Muxed VC1 and E-AC3 to M2TS using tsMuxer It's not working. Can you please post a sample of the extracted VC1 from step 2?

I'n not seeing a sequence layer in the m2ts.

Anyway, who cares about HD-DVD? :)

[woah!]

It's not working. Can you please post a sample of the extracted VC1 from step 2?

I'n not seeing a sequence layer in the m2ts.

Anyway, who cares about HD-DVD? :)

could it be because eacto removes the pulldown from hddvd vc1 files i think...

i suppose you could call your program DGblurayVC1DecNV :D

heres 3 files that should work:

http://tinyurl.com/5c334x
http://tinyurl.com/54uxnt
http://tinyurl.com/67nobk

[Guest]

could it be because eacto removes the pulldown from hddvd vc1 files i think... Nah. The sequence layer is missing it seems. Or maybe it's simple or main profile VC1 in the EVO. That's why I want to see the demuxed ES.

i suppose you could call your program DGblurayVC1DecNV :D Nah. It should work on all advanced profile VC1. Simple and main require stuff to be signaled out-of-band in the container, and if the demuxer doesn't recreate that as in-band syntax, then it goes to hell. At least that's my understanding. Again, that's why I'd like to see the demuxed ES.

Thanks for the samples. I'll report back only if they don't work.

[rack04]

It's not working. Can you please post a sample of the extracted VC1 from step 2?

I'n not seeing a sequence layer in the m2ts.

Anyway, who cares about HD-DVD? :)

I do. :D

http://www.megaupload.com/?d=TL24T9X2

[Guest]

Sharing is caring.

Standby...investigating.

[Guest]

The VC1 elementary stream plays fine, so apparently tsmuxer messed it up.

Are you able to play the m2ts in anything? I can't.

[Guest]

Open your m2ts in tsmuxer and demux the VC1. The demuxed VC1 doesn't play in anything. So it looks like something is going wrong in the muxing.

EDIT: I did a hex comparison of the raw VC1 you gave me and the raw VC1 demuxed from your M2TS. The latter has many spurious 00 00 01 0D start codes (frame start code) inserted throughout the file. It's garbage. You may want to report it to the author.

[rack04]

Open your m2ts in tsmuxer and demux the VC1. The demuxed VC1 doesn't play in anything. So it looks like something is going wrong in the muxing.

EDIT: I did a hex comparison of the raw VC1 you gave me and the raw VC1 demuxed from your M2TS. The latter has many spurious 00 00 01 0D start codes (frame start code) inserted throughout the file. It's garbage. You may want to report it to the author.

I agree that tsmuxer is the problem. Thanks.

EDIT: Submitted this as a bug report to tsmuxer. Thanks.

[bob0r]

Just a side note: there are many cases where tsmuxer creates bad files, and where tsremux does not.
Such cases are:
itv-hd .ts to new .ts
some bbc-hd .ts to new .ts
The new created tsmuxer .ts files are usually desynced.

I never understood this: Why are there so many ways to create, mux, demux or edit .ts files?
For example itv-hd.ts mpc internal, haali, fail to open the .ts, while the muxers and dgavc do open the .ts file.

Guess thats even more reason to support .mkv, it works or it does not.... ok sorry for the off topic, but especially the first part may be useful.

[CruNcher]

I guess because all of the broadcast related stuff like PID, PAT and all the other things you need careful worked out parser logic for that (what do you parse?,when do you parse?,in which situations do you override bitstream values with container level values?) :)

[Guest]

in which situations do you override bitstream values with container level values?) There is no container for transport streams, there is only the bitstream.

[CruNcher]

Of course you right, for transport stream every Video related decision is left @ the bitstream level and it's flags, it would only count for the other most used formats :)
Maybe in bob0r example the problem of not being able to parse those files is caused by missing PAT/PMT or not able to find the correct PID ?

[Guest]

Maybe in bob0r example the problem of not being able to parse those files is caused by missing PAT/PMT or not able to find the correct PID ? Could be. If he would post one, I could determine that.

[Sagekilla]

@neuron2: Is there any possibility of a merge of DGVC1DevNV and DGAVCDecNV into one program, possibly when both are more mature?

[Audionut]

rambo2.split.49.ts (http://www.greentreefrog.net.au/~lionel.davey/rambo2.split.49.ts)

[Guest]

@neuron2: Is there any possibility of a merge of DGVC1DevNV and DGAVCDecNV into one program, possibly when both are more mature? It's possible but I'll decide that later.

[liquidator87]

Might be a stupid question, but will I be able to use this tool w/o a nVidia card?

[Audionut]

Might be a stupid question, but will I be able to use this tool w/o a nVidia card?


From the first post in this thread.

For those not acquainted with my tools, the NV on the end denotes a version that runs on an Nvidia graphics card GPU.

[liquidator87]

Soooo.... well, another question: assuming I've an nVidia card, any chance to get this tool on linux?

[Guest]

Avisynth does not run under Linux.

[liquidator87]

Well in some ways... it does. Using avs2yuv and wine, I'm able to make encodes, if the source is something that does not depend from DirectShow (which obviously doesn't exist on linux).
For example, DGAVCDec and DGMPGDec are great for this, but with VC1 there's not a reliable source filter yet, since FFMpegSource is still young and DirectShowSource can't work on linux...

[Doom9]

That's not really Linux.. Wine is a Windows Emulator so it's basically Windows. And add to that the added complexity of low level hardware drivers that you need to wrap around to - apparently some have done it (http://forums.nvidia.com/lofiversion/index.php?t71784.html) but I have my doubts that you'll get any kind of support for such a hack from either nVidia or neuron2. There's a reason why nVidia just recently released VDPAU..

And I think it's time to get back on topic.

[Guest]

CUDA video decoding is currently tied to D3D so porting to Linux is not an option.

DGVC1IndexNV is finished. I hope to have the source filter done by Monday. Then I'll make a release. BTW, existing licenses will work for this and any future closed tools I may make.

[Guest]

Here's a pre-alpha of DGVC1IndexNV. Your existing DGAVCDecNV license file can be used.

I'd appreciate any test results. I'm starting the DGVC1DecodeNV component now.

http://neuron2.net/misc/DGVC1IndexNV.zip

It supports Advanced profile ES, TS, and M2TS. Please update nvcuvid.dll with the included one.

[Audionut]

Working fine here with the few streams I have.

I've got some more VC1 blu-ray on their way.

[bigotti5]

Should 'Honor Pulldown Flags' work?
It does not here...

[Guest]

Not yet implemented. May I please have your stream for testing?

[woah!]

with the 00006.m2ts file i linked earlier too, i get a error 700 message with it. not with the other 2 tho...



http://img520.imageshack.us/img520/236/error700mg9.jpg (http://imageshack.us)


and here another clip for you..

http://tinyurl.com/6bsdt4

[bigotti5]

Here (http://rapidshare.com/files/167805230/PAL_NTSC_IL_VC1.rar.html) are real interlaced samples, PAL and NTSC

Pulldowned stream is only a selfmade bars sample

[Guest]

with the 00006.m2ts file i linked earlier too, i get a error 700 message with it. not with the other 2 tho... I can't duplicate that. Does it happen at the beginning or end?

and here another clip for you..

http://tinyurl.com/6bsdt4 Thanks but I don't know what to do with an AVI.

[Guest]

Here are real interlaced samples, PAL and NTSC Thank you.

Pulldowned stream is only a selfmade bars sample How did you make it?

[Audionut]

Windows Media 9 Advanced Profile - aka VC1 - Encoded by

Adobe Media Encoder CS4.ts (http://www.greentreefrog.net.au/~lionel.davey/Adobe%20Media%20Encoder%20CS4.ts)
Sony Vegas Pro 8.ts (http://www.greentreefrog.net.au/~lionel.davey/Sony%20Vegas%20Pro%208.ts)

[Audionut]

batman.split.13.ts (http://www.greentreefrog.net.au/~lionel.davey/batman.split.13.ts)

[woah!]

I can't duplicate that. Does it happen at the beginning or end?

Thanks but I don't know what to do with an AVI.

at the beginning but after that it previews the file fine.

that avi? was cut using mkv, then demuxed to the raw vc1 or so i thought. it plays fine through your program tho heh..

what is the best way to cut a clip from a m2ts file then?

[Audionut]

ATM for VC1 i'm using this.

http://forum.doom9.org/showthread.php?p=1196166#post1196166

It can load mkv files and save to .ts or .m2ts which is fine for DGVC1DecNV.
It can also demux from .ts or .m2ts and split files.

[woah!]

ATM for VC1 i'm using this.

http://forum.doom9.org/showthread.php?p=1196166#post1196166

It can load mkv files and save to .ts or .m2ts which is fine for DGVC1DecNV.
It can also demux from .ts or .m2ts and split files.


thx i was doing mkv that way, but had slow playback of the resulting clips? but by doing what ACrowley said here:

http://forum.doom9.org/showthread.php?p=1197472#post1197472

and then process the vc1 dumped file to tsmuxer, cutting the vc1 file in m2ts clips i have no slow playback and the files are attached here to test (hopefully not avi)

split.34.m2ts from a hddvd hotfuzz movie has a pulldown showing in tsmuxer, so here are the same files, 1 with pulldown left in, and the other with tsmuxer removing pulldown:


pulldown: http://tinyurl.com/6f3wmp
pulldown removed: http://tinyurl.com/5n4nb5

another clip no pulldown: http://tinyurl.com/6gb3le

all 3 seems to play ok in DGVC1DecNV...

[woah!]

also i get this alot in all the files i have processed so far, even the 3 i originally posted which are untouched straight off the disc:

SEQ 43151838 !INTERLACE !PULLDOWN
ENTRY 43151860 !BROKEN !CLOSED

dont know if that is normal?

[Guest]

dont know if that is normal? It's just indexing the file positions of the sequence header and entry point and values of some of their flags.

[Sharc]

Pre-alpha works here in preview and play mode, - tested with PAL BD

Edit:
Audio tracks (6 streams) are not demuxed when saving a project, only .dgv is output like:

STREAM 1
PKTSIZ 192
VPID 4113
FO 2
YUVRGB 1
LUMFILT 0 0
CLIP 0 0 0 0
RANGE 0 21613498367
AUDIO 1100,1101,1102,1103,1104,1105,1106

SEQ 990 !INTERLACE PULLDOWN
ENTRY 1012 !BROKEN CLOSED
FRM I PROG
FRM Pskip PROG
FRM B PROG
FRM B PROG
FRM Pskip PROG
FRM B PROG
FRM B PROG
FRM Pskip PROG
FRM B PROG
FRM B PROG
FRM Pskip PROG
FRM B PROG
FRM B PROG
FRM Pskip PROG
SEQ 175326 !INTERLACE PULLDOWN
ENTRY 175348 !BROKEN !CLOSED
FRM I PROG
FRM B PROG
FRM B PROG
FRM Pskip PROG
FRM B PROG
etc.