Help improving VHS captures with Avisynth, capture card?
 

So a few years ago I decided I would transfer around 100 hours of old VHS recordings to my PC. I already knew a little about programs like Avisynth and VirtualDub. I learned some basic things about improving my captures. I had to set aside the project for a while and now I've had to re-learn a few things and maybe finally begin to get this done.

Right off let me say that I know my set-up is far from ideal. Most were recorded by me with VCR's connected to a cable box or even antenna 25 years ago or so when I was just a kid. I have a cheap Roxio capture device and a basic JVC HR-A591U VCR. That will just have to do. It is what it is. I'm actually pretty happy with the final results of few test captures, but there are a couple of things I would like to improve. I know I could combine the scripts, but I have an old/slow PC and don't want to leave it tied up all day, so I split it into steps.

Here are my steps:

I use Graph Edit, VDub, and an Avisynth script to get the levels as close as I can.

Capture with AmaRecTV because I could never get audio sync with VDub. AmaRecTV does perfect.

With my first script, I use AutoAdjust and FixVHSOversharp. I know some don't like automatic filters like AutoAdjust, but it does good enough for me after adjusting the parameters. I don't have time to go scene by scene and make manual adjustments. I have to use VDub 32 bit because FixVHSOversharp is only 32 bit. Is there a 64 bit alternative?

AutoAdjust Script:

Next I use a VHS denoise script that is usually recommended. I made some slight changes.

VHS Denoise

Now I deinterlace with QTGMC. Yes, I know some would say never do it, but I like the results and it will only be stored/played on a PC and sometimes streamed to a TV.

QTGMC:

Finally, I do a final levels adjustment and color balance. Maybe not necessary? I don't know.

Final Adjust:

The only things that still bugs me is the double image. A lot of the captures have it. This may be one of the more severe. Anything that can be done with Avisynth/VDub? Some have that double image or halo all the way around objects. I know a "trained eye" may be able to see other problems and I know that the experts on this board would say get better gear. That is not happening unfortunately. I'm open to suggestions on how to improve with what I have.

"Capture.avi" is the capture(obviously) and "Final.avi" is the result of the scripts. I would probably then be converting to h.264 or h.265 to save space. Thanks for the help.

Capture: https://www.dropbox.com/s/2m05tjjv3k...pture.avi?dl=0
Final: https://www.dropbox.com/s/g5tg7iyx3h...Final.avi?dl=0

That "caprture" doesn't look like an original capture to me.
The edge ghost isn't oversharpning. How do you live with it?

It's the original capture. Now I don't know if the VHS is the original recording or if it is a copy. I recorded these when I was a kid. This ghosting is probably the worse that I've seen with my test. Most others are not as severe. Is there anything that can be done to at least minimize it?

-- merged --

Would this Diamond ATI TV Wonder HD 650 be better? I know it is listed as one of the recommended cards but also can be issues installing it. I have an older AMD dual core with Windows 10. It has an empty pcie slot.

No. The ATI 650 (both USB and PCI) have serious AGC issues. Your image randomly gets darker and brighter -- and it has nothing to do with Macrovision. It's just a really badly designed card, flawed chipset.

Roxio rebadged cheap EZcap/Easycrap (infamously known as "Easycrap") devices. Terrible as well.

There are several options for quality USB capture.
And then I still have a few left as well: http://www.digitalfaq.com/forum/mark...ati-600-a.html

At very least, get a better capture card. The hallmark os a low-end device is color compression, exaggerated chroma noise, "cooked" color (harsh contrast, artifacts).

However, a better VCR will also help tremendously. It can resolve issues that you're trying to fight in Avisynth, either by reducing the problem or removing it entirely. There is chroma noise all over the place, as well as timing wiggles in the image. The wiggling is better than average, but still present -- and it cannot at all be removed in software, reduced by aggressive temporal NR at most.

Even after using Avisynth, you have timing flicker -- probably BOTH from the EasyCap, as well as the VCR.

Overall, not terrible. I see much worse stuff. But it could be better, with better hardware in the pre-conversion workflow.

The double image and blurriness can be inside the VCR, as well as caused by the capture card. This issues probably exists on the tape, both visually and embedded in the signal, but it can easily get enhanced with not-great stuff powering the playback and capture.

This is one of those instances where "you get what you pay for" is 100% accurate.

Thanks for the response.

I will not be spending $75 on a capture card right now. I don't have much income coming in right now. I will look on eBay and take my chances. Would the ATI 600 pcie card be good or is just the usb version recommended for my setup? What are some of the other recommended devices in my case? I forgot to mention it is Windows 10 64 bit.

Some of the videos may never even be watched again. I just want to save them in case I want to watch/show kids one day.

If I could get rid of or at least reduce the double image and timing wiggles, then I would be mostly happy. Better noise reduction, levels, and color would be a nice bonus.

If that's the main focus, then ignore the new capture cards for now.

The timing wiggles have to come from a the VCR. A line TBC is required to correct that, and nothing can be done in software -- aside from maybe reduce it by use harsh temporal NR.

The double image is again also probably (in part) exacerbated by the VCR, though present on the tape itself. For that, Avisynth filtering is your only choice.

So, if you spend any money, put it towards a VCR. See the suggested VCR guide here, know that I have a few available, and you're welcome to take a gamble for something on eBay (and I emphasize the word "gamble" when it comes to eBay VCRs).

The low quality capture device you are using is clipping blacks. It's obvious in the dark areas of your capture and in histograms. Using auto agc filters to heighten the darks simply makes the clipped area brighter; it doesn't retrieve any clipped detail or add anything new, it just makes murky darks out of shadow areas, where increased noise from clipping calls for stronger denoising. Because clipping limits dynamic range, the auto filter has your final.avi looking foggy and over exposed. The filter is also blowing away the midrange and making facial details disappear, and large bright areas like the playing card have a glaring effect (there is also some mild bright clipping in the original capture). Final.avi is an example of why no one recommends one-size-fits-all processing setups. They usually work out to be "one-size-fits-none". You can use auto adjust if you want, but in this case you need to do some more manual tuning concerning gamma and low-end expansion.m I've had to use the same filter to correct for seriously bad backlighting in some home movies, but all filters for that purpose require some fine tuning for each situation -- which is why those filters have so many adjustment parameters to begin with. They're not as "automatic" as the names imply.

The ATI 600 USB clips blacks at y=16 in a manner similar to your current card, and experiments show that the clipping occurs ahead of the proc amp controls. I'd stay away from eBay for capture cards, and that caution has been posted here and in other forums thousands of times. You can still find a decent working DMR-ES10 or ES15 for tbc pass-thru, but capture cards and VCRs at eBay are a high-odds gamble. For less than $40 retail you can get a genuine brand new Diamond VC500 USB, a very popular card for many years, without gambling on cheap look-alike no-name copies. The VC500 has been selling on Amazon at nearly half-price now for several years.

No one has a decent anti-ghost filter. The best was years ago with the TMPGenc Plus 2.5 encoder, and even that wasn't perfect. The nearest thing to that filter is Virtualdub's exorcist.vdf, with two simple adjustments that work in a way similar to the older filter. Exorcist leaves something to be desired, but it beats Avisynth's port of similar filters. So while it's not 100% it's better than nothing. Exorcist was posted earlier with a group of other VirtualDub filters, some of which you may already have. The package includes exorcist, gradation curves, ColorMill, Hue-Saturation-Intensity, and Color Camcorder Denoise. The package is virtualdub_filters.zip.

For wiggles, ripples, and other scanline timing errors, the only fix is a built-in line-level tbc or a line tbc pass-thru. There is no other way.

As for color balance I don't see anything wrong with what you've posted. In fact, color looks pretty good from that player. Maybe you need to calibrate your monitor.

I'm going to look more into the VC500 since it doesn't clip blacks. Does the VC500 or ATI 600 USB have a 3D comb filter?

I've tested the ATI 600 USB several times, deeply analyzing the color, and do not see any clipped blacks.

If anything, the levels of the VC500 were about +5 too dark, though easily adjusted with VirtualDub proc amp settings. Most cards need proc amp adjustments anyway. I'm more inclined to believe that something in sanlyn's setup is reacting badly when that card is in his workflow. And that happens, even to me. This has always been something I've wanted to look into more, but I just cannot repeat it on my end. I have nothing against the VC500, and it would probably be my choice if several others were completely unavailable. I'm not dissuaded much by price, quality is my target. Price on the VC500 is just under $50 after tax/shipping, and that's probably the lowest cost not-crap card you'll find.

Anyway, the Easycap/EZcap (and Roxio/others rebadges) do clip blacks. That is very apparent to me as well, in the attached sample clips. They've always been known to cook color, contrast, IRE, etc. Terrible cheap Chinese card. But it's not horrible to the point where I'll tell you that you must spend money you seemingly don't have -- especially when you're more concerned about the issue that need a better VCR. So, for that reason, I suggested you put money towards the better VCR.

You have several things to think about here. :)

Indon't want tio have to go through all the posts Ive seen at digitalfaq and videohelp that clearly shows the ATI 600 clipping at y=16, but here's one thread with histograms and images:
http://www.digitalfaq.com/forum/vide...html#post53199

And another: http://www.digitalfaq.com/forum/vide...html#post51319,
whose examples came from here: http://www.digitalfaq.com/forum/vide...html#post43950,
and here: http://www.digitalfaq.com/forum/vide...html#post43771

Didn't have time today to trundle over to videohelp to find vaporeon800's ATI 600 tests.

The Hauppauge USB Live-2 does the same thing.

Compare black levels with 9600XT and VC500: http://www.digitalfaq.com/forum/vide...html#post43951

Another problem, a reality of VHS, is that blacks were already clipped on the originating signal. So it's not actually present in the latter tape or workflow at all.

I find the opposite is true of VC500, which is crushing blacks, because the levels are too dark by default. Proc amp adjust makes it fine. But I suspect this varies heavily from system to system, as I've seen oddities with other hardware. Same USB capture card, different levels required per system, both monitors calibrated. The only explanation I have is that it's drivers for both the capture and graphics card.

This gets more fun when output is TV vs. computer watching. :sick:

This is all due to NTSC, where the black floor is 16. Clipping is taking 0-15 and making it all 16, which can lose detail, sometimes.

This all said, my main capture setup is ATI AIW anyway. :)

Rather than blame NTSC, computers, monitors, proc amps, and whatnot, I'm afraid I have to depend on the histograms, my eyesight, the collective experience of others, and objective tests I've seen. I returned my ATI 600 years ago for the reason I've outlined here.

ATI All In Wonder since Windows 3.1 and still going strong. VC500 since Windows 7. The sample VC500 captures in the links provided above show no sign of clipping at y=16. They output continuous data all the way to y=0. You'll never get data below y=16 with an ATI 600. and the cutoff occurs before the signal passes thru the software's proc amp controls. Brightening the picture with proc amp controls only raises the clipping point to higher gray values -- which has also been demonstrated in other threads at digitalfaq. To overcome the cutoff from sources or players that output darker than normal signals you need an external proc amp in front of the ATI 600.

I need to learn more about gamma adjustment. Where is some good info?

Learning about gamma in general will probably help. The reason I first learned about gamma, some 20+ years ago, was because of Mac monitors. Wikipedia is always a good place for a primer.

Remember to look for SD analog VHS info second, not HD/etc.

I wish we had more on this topic on this site, but do not -- though a quick forum search is probably worth a check. I don't remember everything we have here.

Just received a VC500 that I got for about $10 on eBay. It's basically brand new. In a quick test I did, it seems to be clipping at 16. No matter how much I turn down the brightness, the histogram shows clipping at 16. Am I doing something wrong or misunderstanding something?

Let's forget about "like new", which is fiction at eBay, even for $10.

You apparently own the only copy of a VC500 that clips at Y=16 while using VirtualDub capture and saving as YUY2 with lossless compression. Members here are anxious to see a sample of what you're getting.

In the 15 years that I have been ordering stuff off eBay, I've never had a bad experience or received something not exactly as described. Well, there was 1 time, but the guy realized there was a problem before shipping it
refunded my money, and shipped it to me anyway. I guess others have had problems that I don't seem to have.

I'll have to get an example of the clipping later tonight. I don't use VDub to capture. I never could get audio to sync. I use AmaRecTV. I do use VDub to open a Avisynth script that uses a graphedit file, cropping, and histogram to get the levels close before capture. No matter how low I set the brightness, everything clips at 16. I'll explain better and get some examples in a few hours.

Remember that sometimes the source is at fault, not the conversion hardware/process.

I've also never experienced the VC500 clipping at y=16, and I use them quite a lot.

There does appear to be an older version of the VC500 that had a different chipset, so I suppose that could be a possibility.

There is also a VC500CXT version and a VC500 for mac. I think the CXT version has the same chipset but it seems to be a bit broken when using it in linux as opposed to the normal VC500 so it seems there are some minor hardware differences. I have one here I just verified that that one does capture down to y=0 (AKA the red area in virtualdub). I don't know what chipset the VC500 for mac uses, possibly some empia stuff, but I don't think there's a windows driver for it anyhow.

Ok. So I tried a different tape and now it's not clipping.

With the VC500, should I capture at 720 x 480 and then resize to 640 x 480? I'm thinking that the resize should be last right after masking (Crop and Addborder). What resize filter should I use in Avisynth?

Yes, capture 720, resize last after all other processing. :congrats:

Which resize filter is good to use in Avisynth?

The favored resizer has been Spline36Resize. Best compromise between clean detail retention and minimum artifacts. Be very careful resizing interlaced, telecined, or unfiltered video.

Capturing at 720 width gives you more horizontal resolution for later processing.

spline36 is fine:

avisource()
filters.....
spline36(1280,720) # example

I 'm trying to learn more about adjusting levels. I get that luminance values levels should be 16-235 to prevent clipping when expanded to RGB 0-255. I adjust the brightness and contrast before capture. Of course, scene to scene this can change so I get it pretty close. After capture, I need to manually tweak the values. Some scenes may have some pixels below or above 16 and/or below or above 235. Is there anything wrong with doing it this way?

Avisource("Video.avi")
AssumeTFF()
Levels(0,1.0,255,16,235,dither=true,coring=false)

If that is not really correct, then what should I use? Levels? ColorYUV? Tweak? Why? Can Autoadjust be good enough for my purposes with the right parameters? I'm using AVS+ if that matters.

So, as usual, life got in the way and I had to put this on hold for a few months. Time to try again.

I just bought an ES10 for about $50 and it should be here in about 1 week. I have the VC500 and a JVC HR-S3500U SVHS. I know the JVC is not one of the top recommended, but I got it for $5 at the thrift store and I figure it's better than what I had.

The SVHS has a "Video Stabilizer." Should this be on or off? It also has "Video Calibration." Should it be on or off? Is the "Tape Dub Mode" the same as "Edit Mode" that I've read about? If so, should it be on? I will turn the ES10's DNR off.

So I run a S-Video cable and red/white rca audio cables from the SVHS output to the ES10 rear input. Then from the ES10 rear output, run S-Video and audio cables to the VC500. Does this all sound correct?

From what I've been able to gather from reading threads here is that the current recommendations are for Video Calibration to be off. I can't tell from the manual if your VCR has any kind of TBC/NR, but I think not. If it does, turning on Video Stabilizer turns that off. I have a couple tapes that need it turned on for the tape to play properly. I've done test captures with the various options turned on and off, and frankly other than Video Stabilizer I have a very hard time seeing any difference in the captures, but then the tapes I'm using are fairly noise free, they're first generation, etc. Best way is to try the options yourself and see what you see.

Erich

UNDERSTANDING GAMMA CORRECTION (Includes image examples. How gamma affects digital camera encoding is discussed, but take that with a grain of salt and scroll down to pay more attention to the pictures that illustrate gamma correction effects.)

It's good that you're paying attention to signal levels during capture. As for the sample code, no one can advise in detail without a sample of the video levels you're working with. Are you including black borders and head-switching noise in your assessment? Do you really want black borders to be y=16 (they will look gray on TV). Are you using a calibrated monitor to assess the final results?

Those questions can't be answered without a sample of video that you're working with. You might have a scene that requires more or less contrast (that's where ColorYUV or Tweak might come in to play), and you should always assess signal levels after using auto filters like AutoAdjust.

The bigger problem is that your questions imply that you don't yet understand what brightness and contrast adjustments are doing. A sample of what you're trying to correct would make it easier to explain and understand.

Yes. Be sure that you turn off the Panasonic's noise reduction. It creates more problems that it solves and is impossible to correct later. There are better denoisers available in post processing. Anyway, I never cared much for built-in player denoisers, most of which were designed for CRT displays and which at this stage are primitive and obsolete. The advantage of a high-end VCR is not so much its noise reduction, which many users turn off, but mainly that better players don't add more noise and defects to the output.

I keep re-reading the earlier posts but I still don't see why you're resizing. Is it for internet posting?

I'm resizing because I'm capturing old TV recordings that were 4:3. It's my understanding that you should capture at 720x480 and then after cropping and adding borders, to resize to 640x480. Is this wrong?

For denoising and chroma noise, I will be using the VHS Denoise script in the first post. I am pretty happy with the results.

I understand that brightness controls the black point and contrast the white point. You adjust the brightness until all or most the pixels are at or close to 16 and adjust the contrast so that the highest values are at or near 235. Is this wrong?

I will not be manually adjusting scene by scene. I will manually adjust so that overall each capture is close enough. I may then use AutoAdjust to get closer.

My monitor could probably use some calibration, but it is going to have to be by sight. I honestly don't know exactly how to do it. Do I use my graphics card settings? The Win10 calibration? The monitor controls? ICC profiles?

Which one would be better then?

ATI 600 USB, ATI AIW USB or Diamond VC500 USB?

Forget the AIW USB.
It's difficult to tell the difference between the ATI 600 USB and the Diamond VC500 USB. Many can tell the difference right away, especially with a histogram that shows black clipping at y=16 with the 600. But some people don't see that well, and others wouldn't know a histogram from a Popsicle. Both devices are decent performers for lossless capture. Any way you look at it, VHS requires cleanup after capture.

My preference is always AIW if using XP, and I prefer the USB over the PCIe. I prefer the separate audio of AGP/PCI, if using Turtle Beach Santa Cruz, but installing AGP/PCI can be a PITA. The USB is more straight-forward, only 1 driver exists, XP only, and it actually works with laptops via VirtualDub (though MMC is hit-or-miss).

I'm just not a VC500 fan, seen too many weird things. I wonder if it's like some Hauppauge products, where the chip shifted mid-production. Perhaps we need to start cracking open cards that have odd reports and samples?

The 600 is obviously good, as are many (not all) of the clones/near-clones.

Now that I have better equipment, I want to take this one step at a time. I'm using the ES10 for pass-through.

The clip titled "No Adjustment" is a capture without changing the levels. On the "Levels Adjusted" I raised the brightness until the VDub histogram did not show any red on the left side. I used crop so the the black edges did not affect the histogram. I also slightly raised the contrast. Then I captured.

To me, "Levels Adjusted" looks a little too bright. Is it actually correct or am I doing it wrong? I know there are other problems, but I'd like to focus on levels right now if that is a good first step to getting the capture "pretty good."

I'm not going to be as picky as some may be, but I do want it to look better. I have way too many hours of tape to be overly picky.

No Adjustment https://drive.google.com/open?id=16b...PU6jfL45vynV8I
Levels Adjusted https://drive.google.com/open?id=1qb...83WNIKOsCxQGvN

In "No Adjustment" darks are crushed throughout.
The "Adjusted" version has a high gamma that makes the second shot look almost overexposed -- but, my friends, is so common with VHS that it should be standard procedure to have to adjust levels differently for different scenes. Unlike digital video, analog levels vary with waveforms, and they're rarely consistent from moment to moment. The idea behind adjusting levels during capture is to avoid crushed darks and clipped brights that can't be recovered later. Post-processing is the refinement and repair of the many faults of VHS, of which murky shadow rendition is an eternal fault.

There are two shots in the the "Adjusted' video. Both need some tweaking. The first shot of the audience is the one that's too bright. It's an almost-available-light audience shot that could stand to have its gamma and bright contrast lowered a bit. We don'tn really care about seeing that much bright detail in a shot of this type, and it probably looked even darker when broadcast on tv. The second shot needs slightly lower gamma (midtones) and brighter brights (the contrast setting in Tweak() should give the right effect for the bright end).

Does this mean that many scenes in typical VHS projects need different processing for some (but not all) maverick levels and/or color changes? Yes. And the audience shot has a lot of low-exposure camera noise. The second, brighter shot has much less noise. If you used a denoiser that targets the darker shot and used the same denoiser for both shots, the brighter shot will look over-filtered (and it probably will be).

Thanks for the response.

I understand it can change scene to scene, but if I can at least get it so that no scene clips or is too terrible, can I later use something like AutoAdjust or something similar to "automatically" get each scene closer? There is just no way I can spend time adjusting scene by scene for 100's of hours of video. I get that it won't be "as good as possible", but I'm ok with good enough, better than it is, or as close as possible without manually doing scene by scene.

Also, how do I determine if gamma needs adjusting and how much to adjust it?