ffdshow Video Decoder?

[thatdude90210]

Does this work with overlay? On my system, it works with default and VMR9 but produces an error when I select overlay. On my Celeron 2.0 Intel 845GV system, VMR9 plus ffdshow kinda overwhelms the system, producing skip frames and stutters. I like how ffdshow sharpens the picture so I'll stick with default, but I liked the transparent menus of overlay.

[mlbdude]

I can use it with Elecard and Overlay with no issues. It also depends on the kind of deinterlacing you are having the decoder do. I can't use the VMR deinterlacing option of Elecard and ffdshow together so I just use ffdshow to deinterlace. It actually does a better job it seems anyway .

When you say "transparent" and Overlay you mean the omitted backgrounds for items and not the true transparency in VMR9?

[stanger89]

Yeah, all thos screen caps are with DXVA disabled (since ffdshow and nVVPP preclude the use of DXVA).

Just wish I had some more horsepower/memory bandwidth to play with. For some reason I can't really do any of the things I want to do with ffdshow. When I add Denoise3D and unsharp mask, I get badd studder, not to mention throwing in resize or any of the fancy deinterlacing techniques available in ffdshow. Looks like it's getting to be time for an upgrade (this is from a 2.4GHz P4)

[thatdude90210]

I mean the omitted backgrounds overlay menu, that's the one I like.

I know I'm using the Elecard, because I can select the Elecard decoder from the ffdshow icon on the taskbar and change settings that affect the video picture. But somehow on my system, when I select overlay with ffdshow, it spits out an error message saying something like cannot connect to overlay. Wonder if it has something to do with my video, I'm using the onboard Intel Graphic extreme.

[stanger89]

Well, since someone asked about DXVA and I'm getting pretty good at grabbing these screenshots:

nVDVD with DXVA on my Radeon 9500

[DFA]

For what it's worth, as I recall that before inserting the NV-VPP, the codec was operating YUY2 packed format. After inserting NV-VPP, the codec went to the UYVY format. I'm still trying to figure these U-V-Y formating / sampling codes out so don't know what I'm saying. Just an observation.

From the first screen capture that Stanger89 posted showing the nVidia applet with ffdshow inserted, the codec remained in YUY2 format. Again, just an observation with no qualification.

It appears image is equally good with or w/o DXVA when the codec is standing alone. I also thought that DXVA was inhibited when using post processors. Thanks for the cofirmation Stanger89.

DFA

[DFA]

Hi:

More commets:

I did some searching around on the AVS Forum and can say that there is some traffic about NVDVD 3.0 and chroma bug. Some say Elecard has it as well except for ver. 2510. Some say ffdshow should include chroma filter as standard. Some think chroma bug depends on U-V-Y output type (we've seen differently here; same chroma problem for YUY2 or UYVY with post processors inserted).

YUY2 and UYVY should have equal PQ. The difference is in the bit order as I understand it. One person claims that UYVY is less taxing on CPU by 5-10%.

Li On at AVS said in one thread / post:
"And I resolved the NVDVD 3.0 Chroma Bug issue so I'm happy now."
Could find no info on what he did so I PM'd him. Probably something like what Stanger89 did but thought I'd ask anyway.

I'm a little confused why the decoder does not exhibit CUE when no post processing is present but shows it when post processing is present. Seems like some kind of chroma filtering gets bypassed somewhere with post processing.

EDIT: Anyway here are some params for the NVDVD3 codec:

HKLM\SOFTWARE\NVIDIA Corporation\Filters\Video\UOffset
HKLM\SOFTWARE\NVIDIA Corporation\Filters\Video\VOffset
HKLM\SOFTWARE\NVIDIA Corporation\Filters\Video\YOffset

Might this be the same Y offset parameter that Stanger89 changed to -2 in ffdshow? According to RegMon, the NVDVD codec has something like 61 different parameters. Anyone want to see them all?

DFA

[stanger89]

Originally posted by DFA
Hi:

More commets:

I did some searching around on the AVS Forum and can say that there is some traffic about NVDVD 3.0 and chroma bug.

I saw that same thread today (was curious if I was the only one to notice it, like if it had something to do with the version of FWMM I'm running).

I'm a little confused why the decoder does not exhibit CUE when no post processing is present but shows it when post processing is present. Seems like some kind of chroma filtering gets bypassed somewhere with post processing.

Perhaps calling it the chroma bug isn't correct, I'm thinking it's more of just a chroma offset. It's possible that some offset is needed when sending the picture to the renderer and when a post-processor is added it gets done twice (once in the decoder and once in the post processor).

[/B][/QUOTE]EDIT: Anyway here are some params for the NVDVD3 codec:

HKLM\SOFTWARE\NVIDIA Corporation\Filters\Video\UOffset
HKLM\SOFTWARE\NVIDIA Corporation\Filters\Video\VOffset
HKLM\SOFTWARE\NVIDIA Corporation\Filters\Video\YOffset

Might this be the same Y offset parameter that Stanger89 changed to -2 in ffdshow? According to RegMon, the NVDVD codec has something like 61 different parameters. Anyone want to see them all?

DFA [/B][/QUOTE]

Not the same parameter but looks like it should have the same effect, although you'd have to change both UOffset and YOffset probably.

[Narflex]

Interesting thread. I've added those filters to the default list for V2.0 now.

If you want to specify more than one filter on that line use semicolons to delimit them:
vidoframe/additional_video_filters=filtername,inputpin,outputpin;filtername,inputpint,outputpin

[stanger89]

Jeff,

Are we supposed to be able to specify post processors as decoders? I think it would be better if Sage had support for ffdshow similart the way it has support for the dscaler directshow filter. That way we would be able to choose specific decoder, post-processor, and renderer without relying on Directshow's filter building magic.

The other thing about adding these as default filters is that they aren't decoders, they are post processors and I'm still not sure why it works.

Glad to see you having some time to participate on the forums again, keep up the good work.

OK, just thought of something else. It would be nice if you could add the Cineplayer Advanced Audio decoder to the default filter list. A word of caution though, I'm not entirely sure this works for anything but DVD (AC3) audio, I've just re-installed Windows and am in the process of getting everything set up, I'll double check once I get everything installed.
videoframe/additional_audio_filters=Sonic Cinemaster@ DS Advanced Audio Decoder,Audio in,Main Audio Out

[jblust]

Originally posted by Narflex
If you want to specify more than one filter on that line use semicolons to delimit them:
vidoframe/additional_video_filters=filtername,inputpin,outputpin;filtername,inputpint,outputpin

Does this mean that we could use this syntax to chain our mpeg decoder of choice to FFDSHOW, rather than messing with filter merits?

[justme]

Nope, just to add additional filters for possible selection in the UI. Not for chaining them together, or determining the chain's order. Although such a feature would be, too cool.

[xlr8shun]

Originally posted by DFA
Sorry for my ignorance, but what is FWMM?

forceware multimedia (nvdvd3)

[DFA]

Stanger89:

I think you're right. Upsampling occurs within the codec from 4:2:0 to 4:2:2 and by definition is where CUE happens if it's going to happen. Some codecs display CUE in SW mode but not DXVA. You show a clean, no-CUE capture for the NVDVD codec in SW mode when stand alone. So, does not seem to exactly qualify as a CUE issue. Like you said, perhaps better described as a chroma offset issue that needs tuning out. Some codecs may be getting a bad rap because of what is being seen after the post processor. Because it sure LOOKS like CUE.

I would like to tune out what offset is showing with the nVidia VPP. What is the clip you are using? The red collar against black is perfect. I don't have anything handy I can think of. Briefly, what is your procedure for the capture since you seem to have it down? Sorry to put you to task.

EDIT: Oh, and thanks Jeff!! Good to have you drop by.

xlr8shun: Thanks. The light bulb went of a little while later after I asked.

DFA

[stanger89]

Actually it's just a clip out of Inuyasha (I was going to give 100 points to anyone who figured that out). It's just a nice spot without any action so it's easy to grab a screen there, plus the diagonal red on black makes it easy to see.

[xlr8shun]

Originally posted by DFA
If your not getting the nVidia tray icon, I would guess that:

a) The codec is not properly registered or

b) You are not actually using the nVidia codec in Sage

You can re-register the codec by opening a DOS prompt and navigate to the directory where the nVidia codecs reside. Then from the DOS prompt, enter "regsvr32 nvviddec.ax" and after that "regsvr32 nvauddec.ax" and "regsvr32 nvvpp.ax" if you like.

i just noticed, whats happening is the tray icon flashes briefly for a split second when video starts, (or the player is refreshed), i am using the nvidia codec though.

ive tried registering them manually too, again, and again, and again...

in anycase, thanks for the reghack, didnt know about that.. i've had nothing but problems with nvdvd3 codecs.. (guess thats why its beta ) so i'll keep playing around with it, somethings gotta work..

thanks

[DFA]

xlr8shun:

I had trouble getting the codecs to work as well but the trouble was of my own making. I chronicled all that in this thread: http://forums.freytechnologies.com/f...0&pagenumber=2

There is also a line item that can be used to disable the tray icon (dWord <DisableTrayIcon>) but would not know how that would have gotten in the "Video" key because it does not show by default (not the tray icon but the registery parameter). You would have seen it while you were there. Besides, that would have nothing to do with the malfunction.

But you might try what got me out of my self imposed conflict: delete eveything in the NVDVD filter "video" registery key. That will put the codec completely at default. For more detail, read my lengthy post at the end of the referenced thread. Also, I would start out with Sage in overlay as an easy start point. I know your frustration. But really, the codec does work well. The rev. I have is 3.0.50.0.

EDIT: Go to next page, top post for more comment.

DFA

[DFA]

Stanger89:

I found this thread: http://www.avsforum.com/avs-vb/showt...=NVIDIA+chroma

Interestingly, it did not show up in previous searches using the SAME key words. Folks are complaining about the search engine on the AVS Forum and rightfully so. It's flakey.

EDIT: I now noticed he just started the thread and was not there when I first was looking. I suppose he opened it after he received the PM I sent him asking what he had done. Stanger89, I hope you add your solution to the thread.

DFA

[thatdude90210]

Originally posted by thatdude90210
Does this work with overlay? On my system, it works with default and VMR9 but produces an error when I select overlay. On my Celeron 2.0 Intel 845GV system, VMR9 plus ffdshow kinda overwhelms the system, producing skip frames and stutters. I like how ffdshow sharpens the picture so I'll stick with default, but I liked the transparent menus of overlay.

A little update on this problem. I checked the setting on overlay control in ffdshow all the boxes for "supported output colorspaces" and the box for "use overlay mixer" and now it works. I got ffdshow working with my overlay and elecard decoder, nice.

[GbrNole]

for cue to be completely eliminated in nvdvd 3.0 the nv postprocessor should be running.

the other thing to make sure of is that we're not confusing cue with a resize issue.

when resizing with ffdshow both lanczos and spline create faint horizontal lines in deep bold colors and they are much more aparent in cartoon content than regular film content.

however stranger's images were more pronounced in the reds which is consistant with chroma bug.

so i'll quit rambing...