Bought my first capture card

[osgZach]

I recently purchased a Micomsoft SC-512N1-L/DVI for capturing from my PC screen - the main attractor for me was 1920x1080 60FPS support. I'm very happy with the card overall, aside from some minor quibbles, but as its my first capture card I have a question.

One thing I have noticed about captured footage is that it seems to hold up better color-wise. The device natively captures in YUY2 (so, 4:2:2 chroma) and I have noticed that even when capturing in 4:2:0 (x264vfw/ convert to 4:2:0) the colors do not wash out / fade as badly as when I capture directly on my PC in any given program. I can usually spot only very minor differences in contrast to direct captures on the PC done in RGB and then converted (or captured in YV12 via codec options)

I confirmed with another card owner they see very little diference as well. But I'm kind of puzzled about this. I've done everything I can to try and confirm the footage is actually 4:2:0, but FFmpeg seems unable to index the files I produce (H.264 in AVI), and though Avisynth will read them via AviSource it identified them as YV24 but even at that point throwing a YV12 conversion at the footage really doesn't seem to change anything. Vdub seems to believe they are YV12 though (checked with null transform filter).

Is it really that bad to try and go from RGB to YV12 as opposed from doing it from YUY2? Is there something inherently superior going on with the capture hardware? (I think it has a chip that handles color conversion?) I want to think that's the case but my assumption has always been Avisynth / Vdub etc are inherently accurate with their algorithms.

This really perplexes me.

[ChiDragon]

If your graphics card is outputting RGB to the capture card, the cap card may be adding dithering (my AVerMedia cards do this).

Still, I doubt that would help particularly with a 4:2:0 conversion. Have you actually compared difficult content using both methods, or only games?

[osgZach]

I may have to revisit this when I get around to formatting. I haven't had a clean system in ages and I think some stuff is messed up because I get weird results from things now.. Seeing very little difference between 4:2:0 and 4:2:2 in games I know there should be, and getting whack results with certain games as well.

For instance I'm currently scratching my head at testing I've done on the old Battlezone remake. Whether I set x264vfw to capture in 4:2:0 or to retain the color space, I get the same result which is a blurry image with some color loss noted in the UI for sure. If I keep input colorspace the file generated still identifies as 4:2:0, not sure if that's an aberation or not, as its a really old game and may be using a 16-bit display mode. It's really weird because other 4:2:2 captures accurately reflect the colorspace in media info IIRC . Yet funny enough when I play it back in MPC-HC LAV identifies it as YUY2 like it should (but still looks like crap). Load it up in TMPGEnc Video Mastering Works (its really great since they added x264 supported output) and the picture looks much more like I am expecting, better color quality, etc. Almost RGB like.

In short I have too much going on that may be skewing my results, with codecs installed and configuration.. I've always defaulted to using CCCP cause I am too antsy about trying to install a bunch of stuff manually. But I probably need to learn..

On top of that I'm now using a much better calibrated monitor with an sRGB mode (I only tweak gamma via windows and I don't turn on other color modifications from GPU drivers, etc). Although I'm not sure that would account for the drastic difference in how YV12 looks, concerning washed out colors and stuff..

I will re-address this when I've started from a clean system..

[TheSkiller]

Sounds like you are experiencing issues with misinterpreted/mishandeled levels rather than chroma subsampling. The difference between 4:2:2 and 4:2:0 is almost impossible to tell with *most* natural images. With games it can be more apparent if there is, for example, small colored text and such. There should never be a difference where the image appears very different, like much more pale or, on the other hand, more contrasted. This indicates a levels issue.

My advice to find out what's going on: capture some footage in YUY2 (uncompressed or Ut Video 4:2:2) and then examine the captured video's levels and report back. You may try ColorYUV(analyze=true).

[osgZach]

I'm not really sure its a levels issue. But just to clarify, when I sale things look pale or washed out, I'm referring specifically to certain colors, bright greens, reds and blues / aquas that kind of stuff. NOT the entire image. But when I say the videos are always darker (or more contrasted as you put it) I believe that to be because I am used to a brighter gamma value, and with games developers tend to mess with the gamma ramps anyway so it can be hard to actually get a capture that matches what you see on the screen IMHO.. I don't believe there is anything aberrant in the fact my captures are darker, I just fix them in post (although I usually end up tweaking Luminance / Contrast as gamma brightens the darks up but over-saturates the brightness on the brights).

I did 2 captures, they both look like PC levels to me?

(attachment pending approval)