ffmpegsource2 Blank initial frame

[chipzoller]

I'm getting an error when trying to open an MKV VC-1 BluRay stream (2048 x 1080, 29.97) using ffmpegsource2 (v2.16) in avsp: "Insanity detected: decoder returned an empty frame". I've tried changing several parameters including seekmode with no joy. Apparently, ffdshow x64 will not play this stream back via libavcodec but WMVideo decoder DMO in Windows 7 will. This is the first problem I've had opening such streams via FFMS2 and can't figure out what in this stream is causing it to fail. I've even tried re-muxing the video stream to MKV again with no luck. Any ideas?

[TheFluff]

Try ffvidesource("derp.mkv", threads=1). The error message means exactly what it says though; for some reason libavcodec decided to output a frame without any data in it. Before 2.16 the behavior with such frames was just to crash.

[chipzoller]

No change. FWIW, this is a stream extracted from my "Life" box set (Attenborough). Every one has opened without a hitch and was progressive; the last, "Plants", is proving problematic and also interlaced. I've tried extracting the stream several times with eac3to provided with MeGUI (svn 2028).

[TheFluff]

Provide a sample and I'll take a look.

[chipzoller]

Thanks for your help. First 5 seconds here: http://www.mediafire.com/?zp29i622rha0i3i

[TheFluff]

Thanks for your help. First 5 seconds here: http://www.mediafire.com/?zp29i622rha0i3i


[vc1 @ 0x3c220] Interlaced frames/fields support is not implemented


I guess that explains why it doesn't output anything. It also answers the mystery of whether interlaced VC-1 works in FFMS2 or not with a resounding "no".

[chipzoller]

Ah, that would explain it. Any other suggestions as to what I can use as a source plugin/function?

[TheFluff]

dss2() will probably work for your purposes. Another alternative is DGDecodeNV.

[chipzoller]

Oddly, DirectShowSource() and dss2() will open the 5 second sample but not the complete video file (10.9GB). dss() gives me a "can't open" error code 80004005 and DirectShowSource() a timeout waiting for graph to start.

[Guest]

You sample works just fine with DGDecNV. This is the only source filter I know that handles interlaced VC1 correctly.

[chipzoller]

Unfortunately, I don't have an Nvidia GPU or else I would gladly spend the $15 for a license.

[TheRyuu]

Use this (http://fileforum.betanews.com/detail/Win7DSFilterTweaker/1262732488/1) (assuming you're running Windows 7) to make sure that VC1 is using the Microsoft decoders for both 32 and 64bit.

Then try dss2 again.

[chipzoller]

Thanks for the advice but I still get the same error regardless of whether MS or ffdshow is used as the decoder of choice.

[TheFluff]

I find it really odd that it works with the sample but not with the full file. DirectShow graph debugging can be really annoying, but I guess that's what you'll have to start doing. Grab GraphStudio or something and start poking at it; also try with other smaller parts of the file and see what happens.

[Guest]

Unfortunately, I don't have an Nvidia GPU or else I would gladly spend the $15 for a license. You can install a $50 GT520 card alongside your ATI card, assuming you have a PCIe slot available. That card has the VP5 engine, which is the fastest to date for CUVID by a large margin.

[chipzoller]

I might consider doing that in the future, but I'd like to figure this out first. I've never had the need to mess with GraphEdit and so am unsure about how I can construct a graph to have avisynth pick it up for processing. When I render this file, it connects to WMVideo Decoder DMO. If I change VC-1 decoder to wmv9 in ffdshow x64 I can have it renderable in directshow filters with other players, but how can I force that behavior elsewhere? Even when I set the merit to high in ffdshow I can't render the file in the editor.

[TheRyuu]

Can you upload a slightly larger sample that has a reproducible failure (I can take a look at it)?

[chipzoller]

1.5 minute clip here: http://www.mediafire.com/?8adtciod2oy8odb

EDIT: That sample doesn't exhibit the problem I am experiencing. However, I tried splitting larger chunks from the problematic file and found that even dividing it in half produced 2 perfectly fine files. I just combined them back into a whole file and while ffmpegvideosource() still complains, dss2() will render the file now. Really bizarre

[chipzoller]

Thanks again, guys, for offering to help me out with this.