VirtualDub level and histogram settings on captures?
 

I made my first VHS tape captures and I want some feedback if they look good. I've attached three short clips.

My workflow is: JVC HR-S9600U VHS player > DataVideo TBC-4000 time base corrector > ATI All-In-Wonder 9800 Pro capture card connected with 4x AGP > Huffyuv in VirtualDub.

Before capturing I adjusted the brightness and contrast levels on VirtualDub so that each tape's histogram was mostly, if not entirely, in the blue zone with little-to-no spillover in the red zones on the ends. I'm trying to avoid crushing blacks and whites beyond whatever the camera did when it recorded the video in the first place.

Several questions:

1. Should my Huffyuv captures be 35-40GB per hour? I noticed my hour-long capture was 26GB which is much smaller than expected.

2. Should I change the field threshold to 240? Or 480? I think VirtualDub's default setting was 288 which is how I captured the three videos.

3. Other than brightness and contrast levels, should I leave the other levels (color, tint, saturation, sharpness) at the middle of the range?

4. Is it normal NOT to see the overlay screen in real-time while capturing? The image seems laggy or doesn't show anything while successfully capturing, so I've turned it off before capturing.

5. When configuring VirtualDub for capture I looked through the DigitalFAQ guide. I could NOT access Video>Format. I can look under Video>Capture Pin.

The color seems fine, but the contrast may be overdone. It's dark. Not out-of-spec crushed dark, but dark nonetheless.

- clip1 is likely either EP or nth gen, or both. I see the beginning of tearing, chroma noise, and general noise the causes color loss or detail loss
- cliop2 looks to be SP VHS from a camera, and you can see the effects of either the camera, or over-contrast during capture, or both. The light cast onto the chair has no gradient, and jumps from harsh washout (white) to chair color. The kids all hair black hair. I'm not sure if it's okay or not. Hi8/Video tends to have higher contrast like this. Had it not been for the overscan noise, which Hi8 and Video8 usually don't have, I'd have guessed that format and not VHS.
- clip3 is the same as clip2, comment-wise. Note the roses on the table. They have an illegal value, as it's nothing both red chroma on top of white overexposure. Again, not sure if it's the camera or the capture work.

This is one reason why I usually process post-capture. Or use a hardware proc amp, as those aren't as hard as capture software proc amps. Realize this is all nitpicking. Overall, even given the issues, it's within acceptance.

Answers to your questions:

1. Huffyuv compression varies on content. 35-40gb/hour is average. It's plausible to get everything from 25-50. Or maybe even 20-60. It's less concerned about the numbers, and more about quality compression (or lack thereof). It's lossless, so it does not compression the image in any way only the data stream.

2. Interlaced NTSC = x480, PAL = x576. Technically, yes, the interlace threshold is x288, so PAL can record interlaced 352x288, but the DVD players hate it, as it's out of spec non-standard. The usage of this setting is uncommon, but can be done based on needs. I almost never use it. If you capture x288, redo it.

3. The EP/nth VHS had some heavy green shifting, so you can see if capturing with color tweak on proc amp controls helps. If not, leave it alone. Just dial back the contrast some, be sure not to overdarken.

4. In VirtualDub, sometimes 'preview' mode is needed, not 'overlay'. I'm really not sure why, and I've always found it random in behavior. It varies from system to system.

5. I want to update that guide. What are you trying to do?

In agreement with LS here.

A huffyuv threshold of 288 is commonly used for intelaced NTSC and PAL. The threshold is a number of horizontal lines above which huff considers the video as interlaced. With a frame height of 480 lines (NTSC) or 576 lines (PAL), both numbers exceed 288 and are thus treated as interlaced. If you work with old MPEG1 video such as 720x240 lines, 240 is less than 288 and would be treated as progressive -- less than 240-line video can't be interlaced using standard formats. You would set other thresholds for non-standard formats, which you usually won't be capturing anyway.

Agreed Clip1 looks like a generational copy, or at best a tape stored under damaging conditions. All of the clips have most of their darker detail at about a safe y=35 to 50, and drops off below that. But it does vary because of the camera's autogain (AGC) and different lighting. In Clip2 the dim central figures get very slightly brighter as the bright lamp moves into a smaller part of the frame. But the bright lamp causes unlighted areas to go darker because there's no front fill light -- so unlighted people photograph darker than they really were. But at least your camera's meter didn't try to make the figures look like daylight and completely destroy everything on the left side in a huge bright glob. Some post-processing with gamma and midtones without affecting blacks or brights could improve the central figures. Clip 2 has a sharp clipping point in the brights, likely in-camera because of the very bright lamp. Home shots like these with bright lights and darker unlighted areas work OK with your eyes and your camera's viewfinder, but the contrast range far exceeds the capability of most video formats, whether analog or digital. Obviously it's the people who count anyway, unless you just want a video of bright lamps.

In Clip3, when the camera zooms in on the black dog, the dog and the background both brighten (note that the background kitchen counter panel brightens until it almost changes color). Clip3 has some bright overshoot, but not so much that slightly clipped detail can't be partially recovered or tamed. The brightest hot spots like the overhead lamp, the white object on the table at left, and the hot-spot on the girl's hair have no detail anyway.

In Clip1 the girl's hair is very dark blackish cyan, which seems to characterize most of the discolored shadows and darks. In Clip2 the kids' hair is near black but is mostly a very dark brown with a dominance of red, slightly less green, and a smaller portion of blue (that's an RGB proportion that will give you various shades of brown). The hair on the Clip2 girl nearest the camera does have a small amount of detail that a contrast masking filter could bring out if you really want to see more of it but it's probably not worth the trouble.