SmoothVideo Project — LAV video decoder and ffdshow raw filters for 10 bit

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (MAG79) — Thu, 29 Mar 2012 00:41:43 +0000

heyer
May be you are right. It is not important to me, I have not any 10-bit movie.

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (heyer) — Wed, 28 Mar 2012 21:49:50 +0000

MAG79
Well, I don't know how much a screenshot proves, but it can be easily replicated by using ffdshow as the main decoder, enabling SVP and using madVR. On the madVR overlay (ctrl-j) it will show 10-bit input.

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (MAG79) — Mon, 26 Mar 2012 23:45:19 +0000

heyer
madVR reporting 10-bit
Where did you see that? Can you show screenshot?

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (Rimsky) — Mon, 26 Mar 2012 18:21:33 +0000

heyer

thereby wasting CPU
yes, it a waste of CPU time

That seems to be the only explanation to madVR reporting 10-bit, right
maybe, 8bit, 10bit, no matter until mon/TV work in 8bit

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (heyer) — Mon, 26 Mar 2012 16:55:28 +0000

Thanks alot for the explanation
Then I'm guessing ffdshow upconverts it to 10-bit again after avisynth, thereby wasting CPU? That seems to be the only explanation to madVR reporting 10-bit, right?

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (MAG79) — Mon, 26 Mar 2012 15:05:02 +0000

heyer
SVP calculates new frames by Avisynth-script using avisynth SVP libraries.
Avisynth has restriction of color depth. It is 8 bit depth. Avisynth SVP libraries has restriction of color space. It is YV12.
ffdShow used by SVP to get videoframes on-the-fly and to pass it to the avisynth-script. At the avisynth-script input must be YV12 color-space. If not it is converted to YV12 (8 bit depth).

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (heyer) — Mon, 26 Mar 2012 14:53:05 +0000

Btw Rimsky, are you sure about that? MadVR reports 10-bit input when using ffdshow as the decoder, even though SVP is on? The problem only occurs when using ffdshow as a raw filter

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (heyer) — Mon, 26 Mar 2012 13:33:17 +0000

Well, I dunno actually. If it doesn't make a difference, I guess it doesn't matter

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (Chainik) — Mon, 26 Mar 2012 13:28:31 +0000

Do you really see any difference?
10 bit is for encoding quality / file size, not for "i can see 10 bits on my LCD!"

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (Rimsky) — Mon, 26 Mar 2012 13:27:08 +0000

heyer
did you know that AviSynth works only with 8bit pictures? and with SVP all 10-bit video dithers to 8bit becouse it uses avs-script for making interpolation frames?

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (heyer) — Mon, 26 Mar 2012 13:24:21 +0000

Well, the fact that LAV does the dithering instead of madVR, which supposedly is better at it. I really dunno how much you guys can do about it, just thought you might wanna know.

Re: LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (MAG79) — Mon, 26 Mar 2012 12:47:47 +0000

heyer
And what the problem?

LAV video decoder and ffdshow raw filters for 10 bit

null@example.com (heyer) — Mon, 26 Mar 2012 12:42:16 +0000

Hey there, while this issue isn't directly related to you guys, it's still somewhat relevant, since you suggest using LAV for decoding.
When playing back 10-bit files, normally madVR will do the conversion to 8-bit, since its the best at it. However, using ffdshow only for avisynth, it doesn't accept 10-bit input. This means that lav dithers the stream down and not madVR. Using ffdshow for decoding, this issue does not appear.