Ready to start capturing, any last minute suggestions?
 

Hi everyone. I have been on this forum for a little while, and much has been explained to me by posting and lurking the forum. I have what I think is a proper combination of equipment to start converting my family's VHS tapes (50-100 to start) to DVD. However, I want to make sure I am going in the right direction, so please offer any suggestions you have!

Firstly, the capture computer:

Pentium 4 3ghz with HT
2GB RAM
160GB HDD (+1TB external)
Windows XP Professional SP3

On order, I have the following:

JVC SR-V101US SVHS player
AVT-8710
Radeon 9600 All-In-Wonder
quality S-video + audio cables

I also have an older Sony VCR I will be using for dedicated rewinding and fast forwarding of tapes prior to capture.

That is my capture setup. As for editing, encoding, and burning, I will be using my primary computer:

Xeon X5690
24GB RAM
240GB SSD (+2TB Storage)
Windows 7 Ultimate
Avid Media Composer
Adobe Premiere Pro
Sorenson Squeeze

Is there anything you see that I am missing? Still a little unclear on what software to run on the capture computer. Does the all in wonder software come with all I need, or should I be using virtualdub (instead? in addition?) Is there a certain version I should be using?

As for editing, what should I be looking for besides color correction and noise reduction?

I will be using Verbatim printable DVD-R discs (not life/value series)

2Gb that's short (but doable) 4 GB if you can (you'd have to tweak xp i think); disable all programs, antivirus etc..
capture on another (physical) drive, never where there is windows
softs: virtualvcr or vdub

themaster1 is correct, you should capture to a drive that's not the same drive as the Windows OS. If you can install a 2nd drive in the XP computer, that's where you should capture.

2GB for WinXP is OK for capture. XP capture with an ATI 9600 won't use more than 2GB. 3GHz Pent-4 is plenty for ATI.

Some tips:

- Use Virtualdub capture or VirtualVCR. I find Virtualdub easier to work with.
- Start VDub capture with the F6 keyboard key. Use ESC to stop capture. VDub uses the 9600's capture drivers.
- Don't spend too much time on your first test capture. 15 or 20 minutes should be more than enough to reveal any problems or any adjustments you might need.
- Capture to YUY2 lossless compression with huffyuv or Lagarith, 720x480.
- If you have any problems, you can always make a short few seconds of lossless AVI in Virtualdub to post and ask questions. If you need help making a short edit, just ask.

IF by "edit" you mean cut and join operations, use lossless media for that work. You can use any editor you want, but color correction and noise reduction are best done in Avisynth and/or Virtualdub. NLE's aren't really adequate for cleaning up/restoration of VHS -- they weren't designed for restoration or repair. Adobe is good for color work, but be careful about nthe way it converts YUV video to RGB, and use lossless media fort that work. Premiere has a problem with many lossless compressors: you'll do better to ask when it comes time for post-processing.

You'll get much more detail about post-processing if you can submit a sample later.

Welcome. :)

CPU is fine.
RAM is fine.
160gb internal for C: fine, but you shouldn't (maybe cannot) capture to it
1tb internal?

The "never capture to OS drive" rules is from the IDE days. For SATA, it can be fine, if the drive is 7200rpm+ and large enough to have large defragged clean space. If that 1tb external is eSATA, great. That's the second capture drive, assuming the 160gb is IDE (and it probably is, being only a P4). If it's just USB or Firewire, it's not much use to you except to transfer post-capture files.

VCR good.
TBC may be good, watch for faulty chip flicker issues.
ATI card nice, but watch for RFI noise
most s-video cables are quality

FF/REW only deck good idea, save wear-and-tear on JVC

I have the new 6700K CPU, so I wonder who's faster? :P I'll be running some video benchmarks soon.

I don't like Squeeze. The quality has always been... well... squeezed! Is that the new version? I've not yet seen it, but I'm betting MainConcept/Rovi is still better. That includes the Adobe SDK of MC as well.

The ATI AIW card comes with ATI MMC for capturing MPEG. For AVI capturing, you want VirtualDub with the Huffyuv codec installed.

In addition to color tint/cast issues, and various NR, you need to pay attention to chroma noise and chroma offset. That makes pictures blurry and ugly, and can easily be fixed these days. That wasn't always the case.

Verbatim good.

There's really no need for 4gb of RAM on a WinXP capture box. It only sees 3.5gb anyway, and most of it will sit idle at capture. Capturing is all I/O (drive), slight CPU, and barely any RAM.

Attach a small sample to the forum, and we can give feedback if you're unsure.

alright, I got everything set up and took a sample. Let me know what your trained eyes see for room for improvement. This is a baseline capture; everything was left to default once it was set up, the avt is all defaults, VDub is v1.10.4 with huffyyuv codec captured lossless AVI. The audio sounds pretty bad, I'm wondering if my level is just too high for the line in or if that hiss is something else. See attached. The sound card is a Turtle Beach Riviera, but I have had issues with them before, so it could be a crappy sound card.

I never liked the Riviera cards, only the Santa Cruz.

My monitor is IPS and calibrated. The video is almost too dark. I've not metered it, but a quick visual inspection suggests that darks were crushed to 0. Be sure the input settings are not changing the contrast or brightness. Then again, it could just be the tape.

Was this regular Huffyuv compression? For some reason, my laptop is choking on it, which isn't normal.

I don't like VirtualDub 1.10.x, as there are issues with it. It does nothing for capturing or restoration. I prefer 1.9.x. Ain't broke, not need to fix it.

There's definitely hiss, yes. It's a bit higher than normal, so it could be wiring, the audio card itself, the VCR (or VCR settings), or even the tape. If the tape, you'd have to capture more tapes, both commercial and non-commercial, and compare results from them all. If the hiss is identical on all, then you know it's your hardware.

I greatly reduced the hiss by playing with some of the audio settings. I think it was being boosted too much. I agree it doesn't look too pleasing at those levels. There is a massive improvement by bumping up contrast a few points on the avt while dropping brightness one.

I will obtain a copy of 1.9 to try out too and see how that goes. The compression was actually huffyuv 3.0 to 1 I believe. I had set that by mistake rather than lossless. I will retry tonight.

Ps. I was playing a 4 hour tape that has been rewritten multiple times and there was hiss through the entire thing. Is that typical of 4 hour tapes or rewritten tapes?

huffyuv is lossless compression.

[EDIT] As a first quick glance:

Did you capture as RGB? Your video sample is RGB32. You should capture VHS as YUY2 with huffyuv or Lagarith.

The video is telecined (hard 3:2 pulldown). If you plan on deinterlacing (I don't lnow why people insist on it), telecined video shouldn't be deinterlaced. You have to use inverse telecine if you want to restore the original film rate of 23.976 fps progressive.

Looks a little dark, but movies often are. Can be fixed later.

Tapes re-used several times have color and audio corruption.

where would the RGB setting be? I know I selected YUY2 in one set of menus, is there another? Also, can you explain telecine vs inverse telecine? Does it make a difference for home videos?

lastly, are you saying that I should be using the huffyuv (3.0:1 or something) compression, since it is lossless, rather than the 1.0:1 no compression?

in the capture settings> Video> Format> YUY2 (never put rgb) 720x480
huff codec: predict left / predict left (always suggest RGB = Unticked)
And you're good to go

Ok let me ask... Is 720x480 going to be what I want for all tapes, and if not, which would I use 640x480 on? Majority of my tapes are home movies.

720x480 for all NTSC VHS, S-VHS, 8mm, Hi8, and Beta tapes.

640x480 is a computer graphics standard resolution from the good old VGA days.

So I did another test capture (720x480 YUY2 huffyuv). I attached it below. This time I tweaked the AVT as well as contrast/brightness/saturation in VDub. Let me know what you see. I feel like the hue is slightly off (default), but I couldn't get it to look noticeably better so I left it alone. I also tested the audio by capturing with the audio cable to line in disconnected at one point, and sure enough the hiss remained, so I am blaming the sound card. I ordered a replacement.

Thanks for the new sample. Haven't had time to go over it yet. The following notes are for the earlier ca.avi.

My error. As it happened, I used an ancient software app to get statistics for the avi and it erroneously reports some 4:2:2 color codecs as RGB. Not correct. Your ca.AVI is YUY2 with normal huffyuv lossless compression.

Some form of telecine is used to convert movie film frame rates to broadcast standard or DVD/BluRay frame rates. Removing the telecine "combing" effects that you see with 2 of every 5 frames (for NTSC) is called inverse telecine. You can call the removal "reverse telecine" if you like, but inverse telecine is the usual phrase because of methods used to remove the double-image in telecined frames.

Movie cameras that shoot film speed don't apply telecine when shooting. Home camera videos in VHS, Hi8, Dv, etc., are interlaced, not telecined. HD cameras can shoot interlaced or progressive video, not telecined video -- many modern cameras can shoot progressive video at film rates (23.976/24fps).

If you record TV shows with your VCR, be aware that TV shows that originate as film are broadcast (and recorded) as telecined video if you're in NTSC land. In PAL land broadcast is usually interlaced, but many odd forms of pulldown or field blending are applied in PAL country for conversion to NTSC, and vice versa.

Concerning the ca.vi sample:

That's unusual, since most VHS captures would benefit instead by raising brightness and lowering contrast. But that depends on the tape. As it is, your capture is dim, and dark detail is crushed. "Crushed" means that data describing details of dark or near-dark detail no longer exists. Gamma (midtones) are also low, which affects color balance.

It's possible to use filters like contrast masks to pull up some of the dark stuff, but this also reveals the darker crushed areas which, as usual with hard clipping or crushing, contain only noisy garbage. Using an uncalibrated monitor might make your vids look a certain way in your own living room, but they won't look that way to other viewers. Best to use Virtualdub's capture histogram during capture prep and setup to check for video-safe RGB 16-235 input levels. This is difficult to judge by eye alone.

Mild but visible smearing during motion are the bain of players with dnr, which used relatively primitive denoising methods. The worse the noise is on the tape, the more aggressively the filters behave, so noisy and blurry motion are rather typical for many of those players. Some are more blurry than others. Many users turn off dnr during capture, preferring more sophisticated post-process denoising. But again, much depends on the tapes, the player, and the level of original noise.

The glitches mentioned are pretty common with analog capture, even with the best of gear. Aged or poorly recorded tapes don't help. There are other problems that show up less frequently. These bad guys are sometimes easy to fix, sometimes not.

One of the bad guys is shown below in a 2X blowup of part of a frame in the original ca.avi. Notice the noise pattern on the side of the guy's face, in his hair, on the back of his clothing, and in the girl's face. The fine cross-hatch pattern is not grain; some call it FM hash, or dot crawl, or CUE (chroma upsampling error). To me it looks like FM noise. It could be in the cables, the player, or the capture card. If this type of noise exists on all your captures, you have a hardware problem.
http://www.digitalfaq.com/forum/atta...1&d=1449062999

It's possible to smooth crosshatching or dot crawl, as shown below. I used a filter called Checkmate in Avisynth.
http://www.digitalfaq.com/forum/atta...1&d=1449063054

All such filters soften images somewhat, but there are many sharpeners around. Other than using inverse telecine for the mp4 sample attached, this is the only correction I had time to attempt.

Below: less common is bad chroma ghosting, shown here. The image is frame 116 from the original AVI. Pink arrows indicate areas of strong discoloration, which aren't difficult to spot. The color "ghosts" are from earlier frames in the preceding camera shot. The discolorations persist for about 8 frames.
http://www.digitalfaq.com/forum/atta...1&d=1449063133

Compare colors in the above frame 116 with colors in frame 126, 10 frames later, below.
http://www.digitalfaq.com/forum/atta...1&d=1449063191

Below, the bottom half from two original frames, 127 and 130. The upper image is from frame 127. The lowwer image is from frame 130. In this frame sequence, a dark-robed arm is moving upward from the bottom of the frame. On the right lower edge of frame 127 you can see a small part of a white hand as it starts moving up -- by frame 130, the raised hand and robed arm are in full view.
http://www.digitalfaq.com/forum/atta...1&d=1449063246

In the upper image, notice 3 objects in the middle of the picture: the two slightly closed hands of the girl in the middle, the metal tray beside her hands, and the other boy's hand on the edge of the table. In the upper image a pink arrow points to a reddish ghost of the table's wood as the arm is raised and obscures the table.

In the lower image 3 frames later, the raised arm has three main ghost images indicated by the pink arrows: the girl's hands, the metal tray, and the other boy's hand and table edge. These chroma ghosts last for about 4 frames.

Similar effects are visible elsewhere. They are likely the result of sloppy tape production, which is not uncommon (I can testify to that!). But check more than one tape to see if these effects occur elsewhere.

The attached ca_sample_ivtc.mp4 is an inverse telecined version of ca.avi. It's progressive 23.976 fps. Except for removing telecine and smoothing some hash noise, I did no other corrections. Because most smoothing filters are temporal in nature, often the very first and very last frames don't look filtered.

Thanks for the new sample.

The avi is too dark and contrasty, has a chiaroscuro effect. Darks are crushed, and there's hard black clipping. A lot of shadow detail is destroyed. Midtones are suppressed, which affects color perception. Most of the time it looks over saturated, especially the reds. A YUV histogram tells the story. The pink arrow in the histogram below points to crushed blacks and dark detail being clipped in the left-hand unsafe region:

http://www.digitalfaq.com/forum/atta...1&d=1449093369


Not sure which image controls you used in VDub capture. The appropriate filters are in "Video..." -> "Levels". These work in YUV and usually hook into the capture driver's controls. The filters in the "Filters chain" area shouldn't be used, as many work in RGB.

It's not possible to get a perfect VHS capture in all respects. Precise color adjustment during capture is too primitive to be useful. VHS changes color balance every few minutes. Anyway, the signal appears to be too green in the midtones, but not in every shot. You can't make that specific an adjustment during capture without a $2000 studio unit and capturing scenes separately. There are post processing color filters available that are a ton more sophisticated than brightness, contrast, and hue.

The same goes for levels. Analog source changes levels again and again, within a few camera shots. It's far easier to set up a worst-case scenario for a tape, which means adjusting to a suitable range between RGB 16-235 during capture. Set up brightness and contrast to keep the worst scenes from overshooting the spectrum at extremes. Everything can be readjusted later -- that's what post processing is for. Brightness controls black levels. Contrast controls bright levels.

You are using an uncalibrated monitor. Or else, a monitor that's not properly adjusted. The standards for video work are well defined and used everywhere. Try starting at this free website: http://www.lagom.nl/lcd-test/. Free manual test graphics are klunky, but better than nothing.

Let's hope you get it set to your satisfaction. I've been through a few sound cards myself.
:wink2:

I guess I need some help turning all the information you shared into behaviors. I am actually using a calibrated monitor already. I think I need to spend more time practicing though so I can train my eyes about what is correct.

1. Is the chiaroscuro effect a result of too high contrast or too low? Is there a tool (other than my eyes) for me to monitor if my contrast is too high or low?

2. Are crushed darks and black clipping the result of the same cause? Is that because the brightness is off?

3. What is suppressing the midtones?

4. Is there a histogram I can use during capture (or from a paused still) to help set my color/saturation/hue prior to capturing? I tried using the histogram feature within VDub but it did not work. I will try another version when I get home.

5. I was using the levels feature, yes.

6. I am comfortable doing color correction in post (in NLE), I just want to get the best source to work with.

7. "It's far easier to set up a worst-case scenario for a tape, which means adjusting to a suitable range between RGB 16-235 during capture. Set up brightness and contrast to keep the worst scenes from overshooting the spectrum at extremes." Where can I monitor this during capture? Or do I do a test capture and analyze where it falls, adjust, redo, until I get it right, then do the full capture?

8. I ordered an audigy 2 ZS. I'm hoping it is adequate. Suppose it can't be any worse!

Thank you for your patience. There are some things I can call myself an expert in, yet this is not one of them, so I appreciate the expertise you are willing to share. I am fully committed to practicing until I master this task.

I don't see an edit button, couple more questions.

9. In the huffyuv configuration window, YUY2 compression method is "predict median (best)", what should RGB compression method be? It has been on "predict gradient (best)", should it be "convert to YUY2"?

10. under the video drop down, do I want "extend luma black point" as well as white point checked? they are not.

I noticed I had R3 turned on on the JVC. I turned it off now.

I noticed in my video settings (before going to file -> capture avi), my codec settings were not huffyuv YUY2. I changed that to huffyuv YUY2 now. Not sure if it makes a difference outside of the capture settings (which have been huffyuv YUY2).

Figured out historgram only works with preview, not overlay! :smack: I will upload a new capture with the video set in accordance to the histogram.

-- merged --

theres a capture with contrast/brightness set in accordance to histogram, and color +2 on AVT. Default levels for Virtualdub.

FWIW: many older sound cards, especially the generics, had high noise floors. Also, some MB sound sections were noisy. Inside a computer case is a tough environment for audio.

Also, the linear track audio on VHS could hve a high noise floor.

Thanks you for the new capture. Now that we have a better idea of how the tape really looks without numerous enhancements, it becomes clear why you're having a such a struggle with it.

The tape seems to have no contrast, no color density, no sharp or fine detail, does indeed have crushed darks from its production that are not your fault (but settings dark too low makes it worse), has no shadow detail beyond some general murk.....In other words, congratulations. You have what we in this clinically insane activity call a Problem Tape (aka Tape From Hell, etc., etc.).

But you can still work with it. First, let me ask? Is this a retail tape, or is it a tape that was recorded off TV? Some retail taps are actually produced at slow 6-hour speeds (I once had a couple of those. Ghastly). A slow speed tape would offer more clues, especially since you're using a JVC player.

It's getting late here, so I'll have to answer your other 20 questions later. But you did figure out that the histogram works in Preview mode before I could post the solution. Good work. It took me two months to figure that one.

A last point: With black borders in a video, that black border will always show up as a "spike" in the left-hand red area of the histogram. It's usually a small, thin spike, not a sky scraper. The YUV histogram I posted earlier was made with the black borders removed. It's when a stream or even a flood of red shows up in either side that you have dark or bright clipping. You might have noticed when viewing the capture histogram that there was indeed a very tall left-hand spike the size of the Washingon Monument -- that big guy is where blacks got crushed during the production or broadcast of that tape. You might also notice that just to left of that spike along the borderline is a tiny little "tail" that trails off to the left. If you can keep that big fella's tail at the left-hand red side by lowering brightness to keep it close (but not in it), and increase contrast to get just a tad closer to the right-hand bright side (but not too much, or those chapel windows will turn into hot spots when they enter the frame), you'd have a better looking capture.

Take a closer look at a movie when you go to the cinema. Don't trust TV for color lessons, every broadcast will vary. You'll soon get the idea that movie makers seldom go to extremes of dark or bright. To put it simply, the extremes just don't look real.

I'll post more tomorrow. Thanks for your indulgence.

It is a retail tape, and not a very old one. I guess I had higher expectations for a quality capture from this tape because of that, but was definitely wrong to assume that. I am attaching another capture of another retail tape, adjusted for contrast/brightness and slight color boost.(let me know if you think I over-saturated it)

thanks again. The more I play around with it, the more confident I become. The histogram was such a crucial thing. I attached a capture of a retail video as well as a capture of a home movie (from 1988 mind you) so you can see what I am working with.

It's less a matter of "contrast" than it is the absence of midrange content and an emphasis on a wide difference between dark and light objects. Photographically you can get that effect by using a single directional light unit with no fill lights or background lighting. Some people call it "high key" lighting, but that's incorrect -- much of The Wizard Of Oz used high key lighting.

The grayscale image below is a simplified example of chiaroscuro. You can't really say it has "high" contrast, as the brightest pixel in the RGB photo is around RGB 200, well below RGB 235 or 255. Rather, you see brights and darks but not much in between. There is so little detail in the darker shadows that they look as dark as the unlighted background.
http://www.digitalfaq.com/forum/atta...1&d=1449156912

About the histograms above: notice how the YUV and RGB histograms resemble histograms of your earlier captures, with a big spike at the left but not as much data extending to the right. Also notice that in the YUV histogram the spike is contracted at the RGB 16 low-end limit -- but in the RGB histogram at the right, the same spike is at RGB 0. This illustrates that video YUV 16-235 is expanded to 0-255 in RGB display (Seems odd, I know, but that's the way the system works).

If a YUV video histogram extends all the way into the unsafe borders, it's already RGB 0-255, so where will the extreme darks and brights go when expanded? Answer: Nowhere. They get get clipped or crushed. We don't care if a black border is at RGB 0 because that's where a black border is supposed to be. Many TV's don't even display RGB 0-255, as they're designed for a bandwidth of 16-235 for luma and 16-240 for chroma. A lot of TV's actually make black borders look more like "lighter black" (RGB 16). When most TV's try to display RGB 255 or brighter, the brightest brights will glow or exhibit edge ghosting (overshoot). In the YUV histogram the U and V channels (red and blue) are straight lines in the middle: that's because the grayscale photo has no chroma data.

Your eyes are usually the first reference point. If you see that the shadows are nothing more than black gooey blobs, you can be pretty sure your black levels (brightness) are too low. If the blackest blacks or colors look obviously gray or washed out, black levels are too high. IF you see the brightest details as glowing neon hot spots or you see bright details blending and disappearing, brights are being clipped. The tools used to verify what's happening are histograms, waveform monitors, and vectorscopes -- all of which are in Avisynth, Virtualdub, and more advanced NLE's. When elements of a histogram are smashing against the left border, blacks are being crushed. If smashing against the right side or climbing up the walls, brights are clipping. Elements climbing up the border walls are a sign that the video's content exceeds the luma and/or chroma capabilities of the medium.

If you made adjustments with a colorimeter and its software, it should be close to optimal. Pro's advise to work in subdued light and to take a break after 30 minutes by viewing something else (watch TV, read a magazine, etc.). If you watch the same scene for too long a time, your brain compensates for too-dark, too-light, or color imbalance by making things appear "normal".

With reference to contrast, darks are crushed because brightness (black level) is too low. Technically, contrast refers to brights but as most people use it, they are referring to a "contrast range" or a range of brights and darks. So in general layman's terms, high contrast usually means a wide range of dark to light, and low contrast means a narrow range.

It's confusing because many contrast controls don't expand the brights. Rather, they stretch the data from the middle in both directions, making darks darker and brights brighter at the same time, or vice-versa. Some contrast controls come in three flavors: white level contrast affect the brights, black-level contrast affects the darks, and just "contrast" does the stretching or contracting business. Helps to have a histogram to show how they're working, as a lot of NLE's do whatever the designer thought was cool. The Contrast and brightness controls in graphics cards like the ATI's operate according to the standard brightness (black level) and contrast (bright level) arrangement. As you probably found, the two controls interact a bit so you have to do some fiddling. The brightness and contrast controls in expensive PA-100 proc amps and others work the same way. How the same controls in media players work is up for grabs. VLC player's controls are so quirky they're useless.

It's true that a PC displays sRGB 0-255 and can make "TV RGB 16" look a little gray. But media players differ. MPC and MPC-BE expand 16-235 to the way they'd look on TV. VLC Player doesn't. On a properly calibrated monitor in correct low ambient light, you barely notice the difference. Another difference between a PC monitor and a TV is that video looks brighter on a TV because of different luminance and saturation curves. This is why many videophiles don't like watching Hollywood movies on PC's.

With reference to those first movie captures, I'd have to say that the tape is just a very poor transfer. You see an obvious difference between that tape and your other sources. Midtones can also look suppressed with over saturation. The usual saturation control is very general: it saturates everything, lights, mids, darks. Many contrast controls that are the stretching type have a similar effect: the stretching takes mid-level pixels and literally stretches their values both both ways, displacing and lowering the overall amount of midtone data that was right across the middle of the histogram. ColorMill and other advanced image controls let you saturate the three ranges all at the same time or individually.

Do the best you can during capture. Because of the nature of VHS, it won't be perfect. The main idea is to avoid mistakes that can't be corrected later, like hard clipping and crushing. I usually move the tape back and forth, sample a couple of scenes briefly, then set controls for the worst case.

Hue is another story. It's a non-specific control and doesn't work at all like RGB. If you have bluish blacks, green midtones, reddish brights, and over saturated reds, all at the same time (very common with VHS), a Hue control won't fix those problems without screwing up other elements. The PA-100 proc amp has two Hue controls, one each for U and V. Better, but still not as workable as post. Every time I made a hue change in VHS capture I ran into long segments that made me regret it. Whatever you do during capture, you're stuck with it unless you recapture.

Audigy ain't studio quality (neither is Turtle Beach), but it seems to be a decent compromise. I'm using old SB Audigy's from 2004/2006. If you remember that VHS audio isn't so wonderful and varies from tape to tape and player to player, it's a toss-up. I usually audition audio with a Grado SR-80 headphone plugged into the card's output. The best audio I've heard from tapes came from my AG-1980 and my now deceased JVC 8600.

Use "predict median (best)" and "predict gradient (best)". The YUY2 conversion is slower and isn't lossless.

Leave them unchecked. They don't work consistently and can do things you don't want.

Sometimes R3 helps, sometimes not. Depends on the tape. R3 is primitive by today's standards.

Make those settings in the capture app. The two setting groups are used separately.

I made some notes on your latest captures, but I have some chores to finish. Back later.

Yeah, it's rather fiddly at times. But it does give you a leg up on levels.

Working with the Mcqueen cap.avi. It's a good capture with nice color and safe levels. The only glitch was that the black borders and bottom head switching noise affected the histogram. In the avi the black borders come out as RGB 25 or so.So black levels are a little high and images a bit thin looking, but not unworkable.

Below are YUV and RGB histograms from a frame in the McQueen avi. You'd be able to allow for the black borders and let the left end drop down a bit. The RGB has a tiny bit of red overshoot, but that's no problem during post processing. And sometimes you just have to live with such things if they're not extreme.
http://www.digitalfaq.com/forum/atta...1&d=1449185544

This the frame I took those readings from (frame 252 after telecine was removed). It's certai9nly not the horror you had with the earlier tape. It could use some tweaking in the darks below RGB 64 to make it more snappy and realistic.
http://www.digitalfaq.com/forum/atta...1&d=1449185670

Here's the filtered version and its RGB histogram:
http://www.digitalfaq.com/forum/atta...1&d=1449185715

You can see that you were pretty close, even if the borders got in your way. My first step was to shift all pixels to the left (dark side) of the YUV graph by about 6 points (that's about 10 RGB values). That was in Avisynth, where I also cropped the old borders and added new ones to center the image. The main VirtualDub filter I used was gradation curves. In the image below, you can see it in action:

http://www.digitalfaq.com/forum/atta...1&d=1449185886

I moved the diagonal line in that filter into somethging like an "S" curve, which is where the curves filter gets its name. You can use some weird shapes and do crazy stuff, but not here. The bottom of the diagonal line works with darks, the middle part works with midtones, the top part works on brights. Move the line left, it brightens. Move it to the right, it darkens. People pay big bucks for Adobe Pro, Vegas Pro, and After Effects, etc., to get things like a curves filter. The VirtualDub version is free. That filter was tweaked with a second one in the filter chain, then tweaked some more with ColorMill. The RGB histogram is from ColorTools. All these filter thingies look funny at first, but you soon get the idea behind them.

After all that, I got up an mp4 encoded at near DVD or SD-BluRay spec and reinstated the 3:2 pulldown (telecine) to make it 29.97 fps again. Of course you can make it DVD or whatever you want.

These videos have dot crawl and that dratted fine-hair crosshatch. The Checkmate filter cleaned it pretty well. Hopefully you're using s-video cables.

[EDIT] I forgot to attach the DVD-format MPG. Frankly I like MPEG better.

I didn't have time to get into more work on the home video or make more graphics, but I think you'll catch on from the previous post.

This time, the thicker black borders really threw things off, but you did manage to keep everything within bounds. The reason the video looks washed out somewhat is shown in the two YUV histograms below. There's one YUV chart made with the borders still in the video (left). The YUV chart on the right is the result of a histogram with the borders cropped off (right). You can see how the big borders gave you elevated black levels. But the brights are OK. Still a workable capture, though.

http://www.digitalfaq.com/forum/atta...1&d=1449187113

Below, the result of elevated blacks, from the original capture. You can see that borders are still there, at about RGB 18.
http://www.digitalfaq.com/forum/atta...1&d=1449187162

Below, after new borders, a curves filter, and ColorMill:
http://www.digitalfaq.com/forum/atta...1&d=1449187302

Like a lot of home videos, this one is problematic because the camera tries to take in a range of lighting valkus that are broader than the media can manage. You can't have everything, so I decided to let that microwave in the corner disappear a bit (it has very little detail anyway, even at brighter values, and you can see it hasn't turned into a black blob from being crushed with image controls. This was a somewhat quickie correction because I was running out of time tonight. I was less concerned with the microwave and more concerned with keeping detail in mom's and the girl's hair.

This is a tough scene because it has so many very bright and very dark objects, and the glare from the window doesn't help. Get what you can during capture, then fix it in post processing. Trying to capture and do post stuff at the same time is a good way to speed up the aging process, LOL!.

You can crop borders during capture. But be sure to disable cropping before you start capture by setting the crop values back to zero's! I forgot once and left it on, then captured 2 hours at the wrong frame size. Bummer.

awesome. I will have to play around with the RGB histograms in avisynth as well as colormill. I see what you mean about the borders causing the black of the actual video to be light. Adjusting contrast/brightness only with a crop from now on (pre-capture) and it is a great difference (just like the curve corrected). I also got my new sound card today, the audigy 2 ZS and WOW, not only are my speakers sounding way better (clearer, deeper, louder), but my line in noise floor is basically undetectable, just what I wanted. I will be sure to upload some samples this weekend to be critiqued, along with some color mill adjustments to get some feedback on that. Thanks! Have a great weekend.

Yes, with more suitable levels during capture the corrections you'd need won't usually be as extensive as those that I used. But there's always that maverick sequence that comes along.

If you submit filtered videos, we ask that you also submit at least an unprocessed frame capture or two from the same filtered scene. Without knowing what the original looks like, it's difficult (if not downright impossble) to guess what corrections were needed in the fist place. Your newest captures involved mostly levels work. I saw no need to fix any off-color balance. But everyone has their preferences. You'd be surprised at the way appropriate levels make color look more convincing with very little help from the user..

yes I will absolutely post source clips as well. However, I can't seem to get colormill and colortools to work simultaneously. Whenever I open colortools, it takes the spot of the output (or is it input?) window, and then opening colormill, the preview displays the histograms, not the video. Any ideas? I haven't played around with the non-capture portion of vdub so I will try to solve it on my own as well.

Ok, i determined maybe we cannot monitor color histograms with on the fly changes?

Here is a single frame that is uncorrected. I tried exporting as jpeg to desktop but after hitting ok, nothing showed up. So here is an AVI with a single frame. What exactly should I be noting in the colortools and what does it correlate to for colormill?

Write out the colors in their place, to the left in the picture.
Here it looks as if it would be 704 instead of 720th

With a matching Feeder could get a better picture as yet

Yeah, that's one of the klunky "features" of Virtualdub and a lot of free editors that don't have built-in histograms. Also, ColorTools should be kept at the bottom of VDub's filter chain. The filters in he filter list work from top to bottom. Each filter feeds the next filter in turn. So if you have your histogram as the top filter, it won't show you what the other filters are doing.

You can turn off any filter by unchecking the checkbox on the left of each filter in the list. But I guess you know that already.

The preview in VDub filters shows you what is being input to the filter (i.e., if you have other filters preceding it, it also shows you what those filters did). If you make changes in the filter's preview window, the filter's preview does indeed show what the filter is doing -- but you might notice that the Virtualdub output window itself doesn't change until you save the changes and close the filter. Sometimes you have to move back or ahead 1 frame before the VDub output window shows the changes.

Well, to start, your AVI was converted to RGB. That's not a really big problem in this case, but it does show how the conversion blew some bright red off the histogram. Any YUV image converted to RGB will display that kind of problem--that's why maybe 10 YUV frames or more should be in most samples, as many Avisynth example fixes we might show you won't work with a 1-frame video. But sometimes a single frame is enough to show many kinds of problems. In this case the RGB frame does tell us what the RGB colors are doing in display.

To get a single frame from Virtualdub: Use "Video..." -> "Copy source frame to clipboard". Then open a graphics program (you can use Windows Paint) and paste it from the clipboard into the graphics app. That single image gaphic will be RGB. In Paint you can save the output as jpg or png. JPG is lossy compression. PNG is a bigger file but is lossless compression. JPG will usually suffice, but avoid using it for line art like histograms or animation, because it distorts lines and creates artifacts in solid color areas. On the internet many line-art graphics and logos look cleaner because they're usually PNG or GIF.

The filter that Goldwingfahrer shows in the previous post is VDub's chroma shift filter function from FlaXen. I'll post a demo of that chroma problem in a little while.

Goldwingfahrer mentioned 704 width instead of 720. That's another story, but I noticed your earlier home video AVI had side borders. That's normal for 4:3 input, but if you cropped off the side borders and stretched the image to make it 720, it distorts the image proportions horizontally and will look slightly but visibly stretched when displayed.

Goldwingfahrer also mentioned having a better "feed", which I think means a capture with more appropriate black levels. That would likely prevent the red overrun as well as save you some work. But it's best to fix that in YUV first, which is what I did with the other posts and videos.

Back a little later. As usual my home is a madhouse right now.

thank you, stacking the filters in the correct order was the issue and I would have never figured it out.

I converted it to RGB by accident by using a different computer to do the render, and had not yet installed the correct codec :smack:

as for capturing with better black levels, that was with the avt8710 set to not go into the red at any point. From now on I will adjust the levels in virtualdub to get me closer to the boundaries. I just forgot all about those and figured I would have to do it in post. Lesson learned! Same with the 704 vs 720, I punched in to set my levels on the AVT (eliminating the border), and forgot to set it back. It was just a sample capture though, I am much more careful on actual captures. Doing a 2 hour capture as we speak, wish me luck!

I have the Chroma Shift filter only mentioned because I can use it even with YUY2 material.
I do not always need Avisynth after filtering VDub.

Order of the filters is very important to first color shift and then the other filters.

thank you very much :) I am glad you point out such a thing to fix my problem.

But beware, if the script is in Avisynth

ConvertToRGB32(matrix="Rec601",interlaced=false)

Then one must only use the filter in RGB [VDub]
-------------------------------
Although I also have the filter as sanlyn, but I do not take the audio with.

I write
AviSource("G:\daylight.avi",audio = False)#, pixel_type="YUY2")

can sanlyn perfect English ...
listen to him.
Amen.

Chroma ghosting is generally the fault of hardware, not the tape. Look to the VCR or capture card. Probably the VCR. But then again, it can be the tape. You have to really test a workflow with lots of tapes, to get a good overall feel for how it's doing.

I capture as best as I can (VCR, TBC, proc amps, detailers), and tweak in software (VirtualDub, Avisynth, Mercalli, etc).

I don't overkill color correction with histograms, because HDTVs and consumer LCDs vary so much. Those value can be tweaked yet again in that hardware. Eyeball it accurate (on your own color calibrated hardware), but it's doesn't need to be perfection.

Histograms are needed if you're not using calibrated equipment.

True, to a certain extent. Especially if you're experienced and already know what you're doing. IMO If you're just learning color theory and correction, histograms help teach what's going on and aid in developing a skilled eye. Many people can "see" a color problem but can't define it, so they don't really know how to fix it.

The capcm2t.avi can be only partially corrected, and not so very well. The one-frame sample has already been stretched via RGB. Would be easier to fix if levels and saturation could be adjusted first in YUV before doing other work, which is where those problems usually start anyway.

The avi image is below, unchanged. I included some arrows. The upward arrows point to some of the chroma shift (about 4 to 6 pixels to the right). Again, it's more effectively cleaned in YUV. The downward arrow in the lower left corner points to something that might have thrown off your capture histogram: head switching noise can contain very dark and very bright color all by themselves. In this image below, the "black portion" is about RGB 30.
http://www.digitalfaq.com/forum/atta...1&d=1449426894

An inexperienced eye might not notice the neon look of some color areas, especially the dark blue shadows. The grass really doesn't look right, the dark wall in the background is Navy blue, and the skins tones are ugly purple. There's also pink discoloration in different parts of the boy's shirt. Things got complicated because the camera seems to be trying to shoot a strongly side-lighted sun, which glares off the plastic (metal?) toy, and tries to make the unlighted shadows too bright. The scene could have been vastly improved originally by using a polarizing filter, but too late for that. There's also some bright red reflection coming off the red parts. Overall, the image has a blue color cast that affects everything. Any RGB histogram and pixel sampler could give exact values, but an experienced user or a good eye would spot all of it right off.

Newcomers also have a problem understanding hoe to make certain corrections. For example, if a video looks too yellow, where is the "yellow" channel color correction? RGB histograms only show R, G, and B. It helps to know that yellow is made from red and green, and that yellow's opposite, balancing color is blue. Many people try to correct an overly green image by adding red, but that just makes it look yellow. Green's opposite is magenta (Red+blue). Some users are confused by YUV corrections; they look for a Green YUV channel, but there isn't one. It takes some doing to learn that green in YUV is addressed by playing with U and V. If you can look at a YUV or RGB histogram while making changes, you get a better idea of what's going on.

You might want to address the problem of dot crawl. Below, a 2X blowup of part of the above image:
http://www.digitalfaq.com/forum/atta...1&d=1449427194


Below are two RGB histograms. Both were made with the image border present. Both indicate that luma is in a safe range. However, chroma is over saturated. The left-hand RGB is from the uncorrected image. Note that red and blue are climbing up the walls on the bright side. The right-hand RGB is after converting back to YUV and lowering levels and red and blue saturation -- look at how much bright detail was recovered in the right-hand RGB chart.
http://www.digitalfaq.com/forum/atta...1&d=1449427115

The code I used for bottom border, levels and saturation work (in YUY2), plus something for chroma shift that would be more effective with progressive frames or separated fields:
ColorYUV uses a luma offset (off_y=-14) to shift all pixel values toward the darker end, and a negative blue offset to shift blue downward in the same way). The Tweak function lowers saturation, and ChromaShift moves chroma 4 pixels to the left.

Either of those images can be corrected to a certain extent. The version shown below is a correction that uses ColorMill only, with no levels correction in YUV. It looks passablel, but bright red is clipped and some YUV work would make a better image to begin with -- and might not even need much correction at all, had it not been for the blue color cast.
http://www.digitalfaq.com/forum/atta...1&d=1449427394

Below, a levels and border fix followed by ColorMill and gradation curves in VDub. There are differences from the above, mainly because of the same problems in the RGB original. But I don't find either of them to be very convincing color.
http://www.digitalfaq.com/forum/atta...1&d=1449427512

Basically, both corrections are too bright. But just "darkening" would give a very dull image. The attached .zip file cointains two .bcf files. They have the settings I used for the ColorMill-Only version and the Lvels+Vdub version. The VDub filters you need in your VirtuaLDub plugins folder are ColorMill, gradation curves, and ColorTools. Unzip the .zip'd vcf files into a separate folder (your video project folder would be OK. Don't unzip them into your VDub plugins). When you open the video in Virtualdub, use "File..." -> "load processing settings..", navigate to one of the .vcf files, and click OK. The filter(s) will load with the same settings and list order that I used. You can open the filters and examine them to look over the settings used, or change their values in their preview windows and check the affects. Or turn the ColorTools histograms on and off.
Sample_vcfs.zip

Dot crawl is a result of the hardware, and not always of utmost importance to fix. When in motion, it disappear. Not all dot crawl is created equal, or is even horrible. It is what it is. That's just a byproduct of analog video.

I refer to s-video dot crawl, not composite.

It can be seen here in playback, especially during motion, unless something like Checkmate is used to clean it.

here is a capture I am working on. I attached orig and cc(color correction). Please advise on what you would do. This is an old tape of an old film. I have been swamped with finals this week but trying to get this particular tape corrected as best I can to get onto a DVD by Christmas (for the unsuspecting owner of the tape)

Mm. Really bad tape. You might not get many replies on this, as both are huffyuv-MT, which isn't recognized by many media players. Some oddities I saw: (a) both samples are RGB32, which made working with the "original" a hassle, and (b) whatever software you used for your work not only clips darks but hard-clamped them as well and blocked up shadows into grim looking blacks. I don't know how huff-MT is compressing, but when I re-compressed the original to YUY2 with Lagarith I got a file half the size of your sample, and the RGB with the color version was 10MB smaller. That dot crawl is really not good: it's noisy on motion and obscures details. I know the tape isn't so good, but dot crawl that severe makes it worse.

There are three camera shots in the original. The last shot has a totally different color balance -- a nightmare with VHS. I worked the shots separately and joined them to get the non-telecined progressive mp4. Color work isn't as simple as most newcomers think it is, but I don't know how the original looked because of the conversion. Anyway, it'll never look great no matter what anyone does. The girl looks better with white eyes than with orange ones. I did levels in Avisynth and color in VirtualDub. Of course color is largely a personal thingie.

Edit:
Oops, I forgot: Levels in the original you submitted look pretty close to ideal. It's a shame the filtering software spoiled that, but with the original cap intact it's easy to recover.

I thought I had taken the original from virtualdub but I may have run it through premiere pro. As for dot crawl, is that the fault of the vcr, tbc, or capture device?

Both AVI's were made with VirtualDub according to the free MediaInfo analyzer -- but the "cc" version had gone through something else first (PP?), which clamped the low end pretty hard. I'll demonstrate below.

Try a short capture without the AVT8710. That would be my first dot crawl suspect. If not that, I wouldn't suspect the JVC. The ATI card maybe ? ?

I've reduced these sample images to save a little forum bandwidth. The image below is from the "original". Of course it's been converted to RGB when made, so I had to work with that. The black levels and brights are OK. The color is OK, too, maybe needs a little bright blue but not much.
http://www.digitalfaq.com/forum/atta...1&d=1449850469


Below, the same frame from the "CC" version. The YUV histogram shows obvious clamping of darks. Shadow detail is clumped and indistinct, as shown in the histograms. Some dark detail has been wiped out and the image looks unnatural. Brights are OK but look oversharpened, so there's a "clay face" effect and hot specular highlights.
http://www.digitalfaq.com/forum/atta...1&d=1449850507

If you worked levels and color in Premiere you have a leg up on most users, because PP has excellent advanced color tools. PP has all the good measuring tools (histograms) you'll ever need, which you have to learn to use when working with odd video like this. PP doesn't resize, deinterlace, or de-telecine as well as Avisynth, which is the primary tool for that sort of work short of spending 4 to 5 figures on pro studio software. There are similar color tools in After Effects, which I used below.

Below, a frame from the original, unchanged. The "cc" correction looks almost like -- unnatural skin tones, a dull image. It's a maverick shot, much harder to work with than the two shots before it. You can eyeball it and see the low contrast range and red color balance. A pixel reader will confirm what you see: the skin tones need blue, the gal's eyes are yellow, lips are dull. Her teeth are a little reddish in a later frame, but that's chroma bleed (wish we could fix those teeth, but the tape is just too contrary).
http://www.digitalfaq.com/forum/atta...1&d=1449851027

Below, cleaned up and borders centers in Avisynth. I used VirtualDub for color in the mp4, but here I used After Effects. I really wish it was a better tape, but I have plenty of clunkers myself.
http://www.digitalfaq.com/forum/atta...1&d=1449851117

The above image looks sharper, but I used no sharpening on it. Note that you don't need blue in the dark hair, you need it mostly in the midtones and brights. PP's curves filter would be the ideal for that. Could be a bit brighter at the top end, but I ran out of time to tweak forever. I run into a lot of odd color-balanced frames like this in VHS movies. It comes with the territory. And we thought VHS was so great!

It doesn't take long to hang of color work. There are plenty of free Photoshop, Premiere, Vegas, and After Effects tutorial websites dealing specifically with color problems. You don't have to be a pro artist to get the essentials, which are fairly basic.