Scripts and filters for an Opera (sample videos included)

[sanlyn]

pirej
You are right. The way I set NeatVideo is killing too many details and becoming the video too much blurry.

Once again, you're correct: to avoid NeatVideo's default settings. As you've already said, I have small room (352x480) to find a good profile (noise sample) for plugins calculations. It's always saying "area not uniform in Y, Cr, Cb channels", or "area too small". But my biggest mistake was to use its default settings. Your tips on how to find a good place to elect "my noise" were great. Fades and lead-in/out, etc. But a question: that noise sample is good for the whole movie?

I've worked with videos that gave me only a single shot at a sample frame. Yes, it works. It's helpful to have a better source that offers many test areas, so you can see what all those NV settings do. Once you get a filter from a test frame, keep those settings loaded but load up a frame from a different part of your video. Go to NV's preview window and magnify the frame up tp 150 or 200%, then keep changing the Advanced filter settings and hitting Preview to see what they do.

I would thanks god a lot if he gave me a 30 hours day, or even more...

Hmm, in my experience 30 ain't enough. Believe me, even when you think you have it all figured out, along comes a video that tries your patience. Happens every time.

sanlyn, I am not sure you say when talking about grain. Grain I want to remove is that one that is behind the caracters, someway flickering the background. I am not trying to give a digital look to the video, but I'd love to stop that little flickering at background, at curtain of stage. To archive that I put to much blur on filters, killing details, as pirej already said.

Is it flicker, or do you mean a kind of simmering soup of grayish mosquitos and insects flying around against dark backgrounds? That's not grain, that's chroma noise. I should warn you that likely this noise came from the original VCR and was recorded directly to MPEG. Apparently whoever made your copy of this opera did everything you're not supposed to do with VHS. If instead they had used VCR -> AVI -> cleanup -> 720x480-MPEG -> DVD, you'd have more of a fighting chance.

You might try a small cut of your AVI that has some of that noise, and try VirtualDub's built-in "temporal smoother". The buiilt-in is marked ("internal") in the filter list. A value of 5 might be helpful. Over 5, you might start seeing ghosts and other artifacts. Don't use temporal filter in the same run with other filters. Make one run with your temporal filter, save it, then run that filtered clip with your other filters. Temporal filters depend on what happens over a series of multiple frames: if other filters are always changing the sampled frames, the temporal filter's calculations go haywire.

Also, I found a interesting script, maybe some of you already know. I'm talking about a smart guy called Fred. Take a look here:
http://www.vimeo.com/2823934
and here:
http://forum.doom9.org/showthread.php?t=144271

Here you find his script. But as he said at the introduction, his script do not apply very well to low res and VHS footages...

sanlyn, do you think there are something among his script I may use to advantage? If I apply it "as it", result is useless

.

The other participants are much better versed in AviSynth than I am nowadays. I gave up on its filters long ago: I never found an AviSynth filter that could clean chroma noise from VHS or cable transmission.

Hopefully I can take another look at your video samples over the weekend and post something useful. Very hectic these last 4 or 5 weeks.

[sanlyn]

I ripped the video from DVD using DVDDecrypt. After that I used Womble Mpeg to cut some parts and saved it as Mpeg-2 stream. Now, it's show time! How to get a better look of this 3 parts of the video?

http://www.4shared.com/video/ljWcaKt7/forum01.html
http://www.4shared.com/video/4AWcXNJ3/forum02.html
http://www.4shared.com/video/FD8yMg5K/forum03.html

Just had a closer look at these samples. Is it possible for you to post a frame or two from your original master (i.e, before you made any conversions or other processing). You should be able to get a frame capture or two from VirtuaDub. Use any picture viewer to save the capture off the Windows clipboard as uncompressed .bmp -- don't save it as JPG, as most programs make very low-quality JPG's. The video's format is 352x480, so one or two BMP's should not be very large files.

The AVI's have severe color space conversion problems. Have you noticed there's no blue, and that shadows are yellow and/or green? Color correction isn't possible with color space errors. Add all the blue you want to the AVI frames, it will never look correct.

[poisondeathray]

@sanlyn - I believe he posted an original sample "test.mpg" in one of the earlier posts

http://www.4shared.com/video/BKC07Ri6/test.html

[jairovital]

sanlyn
man, you are great! Looking deeply in my samples by an expert is all I'd love to have.

Is it possible for you to post a frame or two from your original master (i.e, before you made any conversions or other processing).

I put more than one or two BMP samples, captured from original VOBs:

http://www.4shared.com/file/Dog6BF9X/samples.html

If you could find the best setting for NeatVideo, or other VD filter, I think I could say my working on Mignon was over (and I admit, a little bit of nightmare would gone...)

Thx
jairo

[davexnet]

Those colors are pretty bad - is that the way it looks when you play the tape and watch on your TV,
or is that something from the capture process ?

[jairovital]

davexnet
I am not sure how it looks on TV. My source is only one DVD, an old one. In fact I explored it and found files I didn't know about:

I:\VIDEO_RM

VIDEO_MN.DAT
VIDEO_MN.IFO
VIDEO_RM.BUP
VIDEO_RM.DAT
VIDEO_RM.IFO

I:\VIDEO_TS

with normal VOBs and IFOs files (each VOB with 650 Mb. Not expected 1 Gb size)

Another weird thing: using PowerDVD I cannot search using tracking line. It isn't work and I can't jump to the middle or end of the video. I don't know why. Video has no chapter at all. It is very annoying if I want to see far from the beginning. Then, I ripped it to HD.

Have you ever heard about VIDEO_RM folder and files?

jairo

[sanlyn]

@sanlyn - I believe he posted an original sample "test.mpg" in one of the earlier posts

http://www.4shared.com/video/BKC07Ri6/test.html

Poison, thank you. I really shouldn't try to work at 2:30 AM. I downloaded that a few days back. My bad.

[jairovital]

sanlyn
don't go to test.mpg. Instead, go to BMPsamples and you'll find many more screenshots from the whole video.

But, for now man, go to bed. 2:30 AM is good for dreaming with codes and scripts...

[sanlyn]

sanlyn
don't go to test.mpg. Instead, go to BMPsamples and you'll find many more screenshots from the whole video.

But, for now man, go to bed. 2:30 AM is good for dreaming with codes and scripts...

I have the BMP's as well, but I still need to see moving video, so I'm looking at both. I see the MPG was made with Womble MPEG DVD Wizard. Is test.mpg cut directly from one of the VOB's, or did it go thru some other extractor first?

[sanlyn]

VIDEO_RM folders appeared only on SVCD's and DVD's created by older Philips DVD recorders. It's not a standard DVD folder; DVD players ignore it.

The reason you can't time-search or see any chapters is because the video is not a DVD. It's a CVD (Chinese version of Super Video -- i.e., a CD, not a DVD). CVD has a maximum bitrate of 2175. My software says this CVD was encoded at 2100 bitrate VBR. This is a low-quality bitrate, designed primarily for PC playback -- not for big-screen TV. You can re-encode at up to 10,000 bits, but it won't improve anything. However, to keep from losing what little quality you have with CVD, you should encode later at 4000 to 4500 VBR.

I don't know what happened to the audio track that was a problem, but GSpot tells me test.mpg has AC3 audio. That could not have been true of the original. CVD used Windows MPEG-Layer-II. While ML2 can carry stereo, its audio bitrate on CVD is usually 40,000, not the 48K DVD standard, so it couldn't have been recorded originally as a VCD with Dolby sound. But I guess someone might have found a way, I don't know.

I opened test.mpg in VirtualDub 1.9.4 with the AC3ACM Dolby translator installed (http://fcchandler.home.comcast.net/~fcchandler/AC3ACM/index.html), and created a huffyuv AVI and separate .WAV PCM file.

Correct me if I'm wrong (OMG, do I have to look foolish this many times in one day?), but is test.mpg from the file that you ripped with DVDDecrypt and extracted with Womble? If so, this is not the media you should be working with. You want to make an AVI directly from the original VOB files. VirtualDub can open and translate VOB's. It can't read AC3 audio unless you also install AC3ACM (see link above). If VirtualDub can open the VOB on that disc, VDub is all you need. Don't try to make an AVI of the entire 3 hours. Use VDub to cut the movie into several AVI segments, saving the audio as PCM .wav (click "File"..."Save WAV...", etc.) or you'll end up waiting hours for a 45-GB AVI. Load and Work small sections at a time. There's no way to clean and convert 3 hours of video in a day.

I was looking over the avi when the late hour caught up with me. Sorry, jairo, I'll resume later. Everyone should have a careful look at what there is to work with.

[poisondeathray]

Just an FYI , virtualdub doesn't handle interlaced MPEG2 correctly , and shouldn't be used (unless you load the MPEG2 file through an avs script and do the RGB interlaced +/- matrix conversion there )

jagabo has noted this in several posts
https://forum.videohelp.com/threads/307116-8mm-film-and-the-msu-old-color-restoration-f...=1#post1890833

DGIndex or DGSplit would be a better tool to cut a native sample from the original . Womble does smart render, but depending on where the GOP was cut, it will re-encode segments

[lordsmurf]

Just an FYI , virtualdub doesn't handle interlaced MPEG2 correctly , and shouldn't be used (unless you load the MPEG2 file through an avs script and do the RGB interlaced +/- matrix conversion there )
jagabo has noted this in several posts
https://forum.videohelp.com/threads/307116-8mm-film-and-the-msu-old-color-restoration-f...=1#post1890833

I just spent 5 hours this morning running experiments. And I'm happy to report that this is not exactly true when the workflow remains interlaced from source to DVD. If you deinterlace, it's over. (The workaround is to pre-encode to AVI in VdubMod, then import that to Vdub for main processing. Workflows may vary, however, depending on how dirty you want to get with Avisynth and indy tools like DGIndex.)

I need/want to run the tests again within the next 3-4 days, for re-verification.

I used jagabo's test clip from another thread, as he'd picked out a good one.

Some of what I saw, I don't really understand. I need to gather a dozen or more screen shots, and then present my findings in a write-up. Coinciding with that, I'll need to open up two new topics to address some questions I saw. I don't have all the answers (and I don't know that any one person will), but a discussion between several of us should produce results on how to approach VirtualDub's odd problem.

What a mess.

I wish the Vdub programmer(s) could just fix this already. (Although I'll admit to not being a programmer at that level, so no idea how hard it is to actually do. Asking is surely easy!

So poisondeathray, jagabo, I think you're wrong ... while at the same time you're right. And I also know why I never saw any issue. .... Assuming the follow-ups give same results.

[sanlyn]

Just an FYI , virtualdub doesn't handle interlaced MPEG2 correctly , and shouldn't be used (unless you load the MPEG2 file through an avs script and do the RGB interlaced +/- matrix conversion there )

jagabo has noted this in several posts
https://forum.videohelp.com/threads/307116-8mm-film-and-the-msu-old-color-restoration-f...=1#post1890833

I just spent 5 hours this morning running experiments. And I'm happy to report that this is not exactly true when the workflow remains interlaced from source to DVD. If you deinterlace, it's over. (The workaround is to pre-encode to AVI in VdubMod, then import that to Vdub for main processing. Workflows may vary, however, depending on how dirty you want to get with Avisynth and indy tools like DGIndex.)

As I said erlier, when working at 230-AM I should save replies in Notepad and wait til next day to post <sigh>. Here's an earlier exchange on the subject with jagabo:
https://www.videohelp.com/forum/archive/new-virtualdub-mpeg-still-a-colorspace-problem-...i-t371664.html

poison & lordsmurf: I'll edit my previous post to jairo concerning MPEG->AVI. Since Feb/2010 I've used VDUb 1.9.4 with FccHandler's MPEG plugin and have not seen the problem again. Examining projects from 2 years ago, I saw it sometimes, but not always. Like LS, I don't know why it was so inconsistent.

But, just in case, and as a compromise to make sure the conversion goes properly, I'll suggest that jairo go with either the DGIndex route or just use VirtualDubMPEG2 1.6.9 (http://home.comcast.net/~fcchandler/stable/ for MPEG->AVI ONLY, not for any other work. VirtualDub should be set to save the AVI as RGB24. NeatVideo works with RGB24/32; internally it uses its own proprietary YUV space, but outputs RGB.

[jairovital]

sanlyn
as we use to say here in Brazil, my humble thread became a *fight of big dogs*.

People like you, davexnet, pirej and now with lordsmurf and poisondeathray are flooding my brain (in a good manner) with too much information my head cpu is running and heating to follow your tips.

But I read you last edited post and I'll try it. Later, I'll post some results.

Keep in touch

[poisondeathray]

@Lord Smurf - It's definitely reproducable , even on an interlaced workflow.

It's more noticeable on anime, but I have some live action sequences where its painfully obvious. You don't even have to pixel peep or zoom in to see something funky is going on. I'll upload a sample clip later

Even if it occurs some of the time (in my opinion it occurs all of the time) , that's enough IMO to spend the extra minute with avisynth.

EDIT: OK a non anime example: top is vdub, bottom is vdubmod, and a sample clip appended below for your own testing. It's easier to see whenever you have red on black or (red on blue) . Make sure you view at full size

[sanlyn]

poisondeathray, I don't at all doubt what you say. I just tried both methods I suggested to jairo above (and I think jagabo and others would approve of both): VirtualDubMPEG2 and VirtualDub/DGIndex/AviSYnth. Don't see any problem either way, and they both look exactly alike. I didn't compare with plain-vanilla VDub 1.9.4 without DGIndex/AviSynth -- I'll save it for further discussion on the issue. For the time being, the other two procedures as suggested displayed no problems.

[poisondeathray]

I know sanlyn, but I think Lord Smurf still has some reservations or wasn't entirely convinced . Sorry for the diversion, carry on... (or get some sleep first)

[sanlyn]

sanlyn
as we use to say here in Brazil, my humble thread became a *fight of big dogs*.

People like you, davexnet, pirej and now with lordsmurf and poisondeathray are flooding my brain (in a good manner) with too much information my head cpu is running and heating to follow your tips.

But I read you last edited post and I'll try it. Later, I'll post some results.

Keep in touch

Just take it one step at a time, jairo. Once a basic procedure gets set, you'll seldom have to change it.

The interlace problem shouldn't occur with either method suggested. VirtualDubMPEG2 is the easiest, but you can use either method. Just don't use VDubMPEG2 for anything other than MPEG->AVI. NeatVideo, AviSynth, and many VDUb filters work only with the latest version of VirtualDub. However, if you do need any color space conversions later for AviSynth procedures, convert with AviSynth but return color to RGB24 with AviSynth when you're finished.

The color on the mpeg test AVI looks more real-world now. Still a very fuzzy image. Will do more later.

[jairovital]

sanlyn
I downloaded AC3ACM codec your suggested and too many good news came: first, I could load original VOBs in VDub 1.9 and hear sound. I used MPEG-2 plugin v3.8 by fccHandler Released February 13, 2010 version. Second: I could hear sound loading original DVD in PowerDVD without have to set audio to Cyberlink Virtual Speaker, as I was doing because Stereo settings wasn't giving me sound. Third, and most strange, now I can seek video back and foward using track line... Three gifts AC3ACM codec gave me.

Now, bad news... I went to VDubMpeg2 v1.6.19 (latest one) and load first VOB to make my AVI. Ok. Saved Wav and all was good. Second VOBs and others showed up with lost of synchronism, sound comming about 1 second after video. Same with VOB2, VOB3,... Only VOB1's sync is correct. No way to carry on with such problem. I think I can chage fps to archive the right sync, but best if I ask you first.

But if I use VDub 1.9 with MPEG-2 plugin v3.8, lip sync is Ok for all VOBs. Can I use VDub 1.9 instead of VDubMpeg2 ? or I need that DGIndex file (D2V) and Avisynth script to do that?

Iam planning to go with huffyuv encoding and separeted Wav route.

If I use D2V and Mpeg2Source command, VDub load no audio... No way to save Wav. Then I loaded AC3 that came from DGIndex, opened it with SoundForge 8, normalized it, filled left mute channel with sound from right channel, saved it as Wav PCM and loaded it in VDub/AudioFromOtherFile. Is it correct?

[jairovital]

Sending reports

01 - I saved DGIndex project that joined all VOBs and splited an AC3 file.
02 - Converted that AC3 to WAV PCM using RiverPastAudioConverter.
03 - Filled left mute channel with right channel sound and normalized it using SoundForge 8.
04 - Opened Avisynth script with Mpeg2Source and ConvertToRGB commands in VDub 1.9 with MPEG-2 plugin v3.8.
05 - Load edited WAV at Audio/FromOtherSource menu.
06 - Made a small cut at beginning and end of the video.
07 - Divided the whole video in pieces of 30,000 frames to output separeted files. Total is 320,000 (!)
08 - Saved WAV for each 30,000's piece.
08 - Huffyuv codec to encode. No Audio checked.
09 - File size of first piece: AVI = 5,4 Gb; WAV = 183 Mb

Before going on, I'll wait gurus and gods say if I am going right or something is wrong.



jairo

[sanlyn]

3 caps from VirtualDub. Original saved as Photoshop images (uncompressed), with no modfication. Converted to JPG 95% quality with Photoshop plugin HVSJPEG-2.

The video was converted to AVI uncompresssed (RGB24). (1) via DGindex/AviSynth. (2) via VirtualDub 9.1.4/FccHandler's MPEG2 plugin. (3)VirtualDubMPEG2 1.6/MPEG2plugin.





I'll be using the DGindex/Avisynth version. They look alike to me. Would still like to know why the Vdub conversion is so inconsistent. Last time I saw a problem was maybe 2 years back. It's been a while since I used MPEG source; I always capture VHS via VirtualDub straight to AVI/RGB24/huffyuv (Never had a problem that way, but there was no MPEG->AVI conversion involved). Will keep looking at the other versions later.

[poisondeathray]

It might not be as clear as in that example I posted above, but the errors are there. It's more clear with motion, and more clear with red on black (or red on blue)

Here is frame #225 by vdub numbering

You can see the highlighted boxes where vdub vs. vdubmod (or avisynth dgindex) handle the interlaced chroma differently. Notice the line spacing on the color bands on top.

The "zoom" was a nearest neighbor resize with mod16 crop on exported png. All captures were done with "copy output frame" Ctrl+2 and pasted into image editor and saved as png.

Imagesource()
Crop(144,208,-112,-160,true)
PointResize(last.width*2,last.height*2)

If you notice in the "zoom" picture, the lines have different color


Anyways this is getting off topic... Let's see what things can be suggested to improve this source

[sanlyn]

It's subtle, poisondeathray, but I see it. Still, all caps show badly serrated edges, no matter how the conversions are made.

You can see from the images above that test.mpg has interlace and color space problems. As with the avi samples earlier, Blue is depressed. However, adding Blue didn't make the overall balance look correct. Below is a Color Tools histogram. All frames in test.mpg give the exact histogram, with depressed Blue. The fault of stage lighting, maybe (tends to be warmer), but yellow shadows in white objects don't look good.




I don't know how the original CVD was made. Originally recorded from a stage production onto VHS? (It doesn't look like a retail VHS). Played back on a VCR with bad heads? Recorded with a low-quality capture device? Recorded directly to a Philips DVD recorder? A cut from an original VOB might look different (wombie having hiccups? Who knows). Maybe we can get a direct piece of a VOB straight thru DGIndex/avi synth to D2v? This won't be easy.

[poisondeathray]

Instead of adding blue, another approach might be to removing green (e.g. one way might be to use RGB curves and lower the the green channel curve). At least with curves you can do non linear changes (e.g. just shadows , or affect highlights less)

But the bmp screenshots posted present other bigger problems. The colors, hue, and levels issues are not consistent between scenes. So I suspect whatever treatment will have to be done differently on different sections. Also the method in which screenshots were taken is important - was the conversion done with RGB full range? Was it Rec601 16-235? etc... Assuming it was consistently done, it still leaves the same issues

I agree this will take a lot of work.

I don't do much of this type of stuff, so I'm trying to learn from you guys as well.

[jairovital]

Paying too much atention what you guys are talking, I got a small piece of m2v file extracted from VOB using DGIndex.

I am not sure if m2v is the same as VOB you were trying to see.

[sanlyn]

Instead of adding blue, another approach might be to removing green (e.g. one way might be to use RGB curves and lower the the green channel curve). At least with curves you can do non linear changes (e.g. just shadows , or affect highlights less).

Actually, there's too much green. The whole video is green, period. I'm mainly using VirtualDub's gradation curve filter, similar to Photoshop's (also using Photoshop to examine frames).

But the bmp screenshots posted present other bigger problems. The colors, hue, and levels issues are not consistent between scenes. So I suspect whatever treatment will have to be done differently on different sections.

Not only do they change from scene to scene, they change within scenes. Look at test.mpg as the camera pans, and you'll see at least 4 major changes in color balance and lighting (I hate those old AGC controls on VCR's). This appears to be a recording opf a stage production, so expect uneven lighting and artsy effects as people move around the stage.

Also the method in which screenshots were taken is important - was the conversion done with RGB full range? Was it Rec601 16-235? etc... Assuming it was consistently done, it still leaves the same issues

I agree this will take a lot of work.

I don't do much of this type of stuff, so I'm trying to learn from you guys as well.

Some work in full-range RGB (PC), some use TV. I think full-range gives a better idea of how video displays on a modern tv. When you see colors jammed up against the left edge of a histogram, you know the color's crushed on PCs and TVs alike. As for 601/709: on a pro monitor ($5000 and up, not counting $25,000 worth of associated gear) you'd have to use that setup, but consumer PC's use RGB. In effect, all 3 standards look very much alike. My monitors are calibrated with Gretag-MacBeth/EyeOne Display2, which lets me set my PC display close to what you'd see on tv. Most PC monitors out of the box are 300x too bright, grossly over-saturated, and have weird gamma curves that look like the Santa Anita interchange.

After all that, you watch the results on TV and you still find more corrections. It never ends.

We're all learning from each other. That's why it's so much F U N

[sanlyn]

Paying too much atention what you guys are talking, I got a small piece of m2v file extracted from VOB using DGIndex.

I am not sure if m2v is the same as VOB you were trying to see.

Thanks, jairo, will take a look. VOB, m2v, D2v, MPG...pretty much the same thing. Sorta.

[sanlyn]

test.mpg caps:
Here are two caps from test.mpg, encoded with TMPGenc (352x480, 4000 VBR bitrate, 5000 max). I think you'll be able to see the effects of making master recordings from VHS at low bitrates and small sizes. Color and image density are a mess, but at least it's not totally green. Blue is crushed in the original, meaning that shadows are always gonna be warmish (red) because blue detail below RGB-40 or so just ain't there. Also note corruption along the right edge. Likely this came from the tape, along with the stripes across the top. No fix for this.



The more people you see, the less you can make out. Highlights are wiped out, as you can see from the white shirts (even tho I set filters to cut contrast above RGB175 or so; the details were wiped out long ago. Note: to the right of the red center figure and behind him, there's a dense brownish shadow. This blob changes shape occasionally as the camera moves, but it's totally crushed: nothing there to retrieve. Meanwhile our fair damsel at the top has gone easter-egg green. You can't make the imager righter; highlights are already toasted and there's very little shadow detail to work on.



I wouldn't use the Wombie method for processing. Interlace and color problems are severe, even worse than plain VirtualDub conversions.

[sanlyn]

Here's a cap from the D2v you posted (thank you). reported the file was not a D2v project file, so I used the m2v instead. This is from the MPEG2 encoded with TMPGenc. AviSynmth/DGIndex Pretty noisy chunk of video, but VDub's built-in temporal smoother set at 5 removed almost all of it, NeatVideo with very low settings killed the rest. The DGIndex cut is more workable, but let's face it -- the small capture size and low bitrate don't give you much to work with. But it still looks somewhat better than the test.mpg video.The original VHS must surely have been of higher quality than the CVD.

I corrected some color, though the scene doesn't clue you as to what it should look like. I assumed the guys' leggings (stockings ? Whatever.) should be a light, generic flesh tone. The shoes and the pirate hat are black, so is the cloth covering the gal in the background. Once I got black for these objects, most colors fell into place. There's no real RGB-000 black in here: again, dark Blue was crushed, i.e., dark blue detail is nonexistent. This is a soft image that's actually too soft to sharpen. I set NeatVideo's sharpener down to 20%; setting it higher gave an etched-glass look that made horrible motion noise on an LCD every time something moved. Getting the colors smooth is impossible: look at the lady standing in red at the upper left. Colors are smeared and tainted with chroma bleed and low-bitrate artifacts.



I'm posting below an example of a workable VHS capture from the movie "Lili", a very old and slightly damaged VHS I've worked on occasionally for 2 years. This scene has been thru NeatVideo, but it needs a ton of color and other work. I post it to show the difference between a direct-to-AVI capture -vs- other ways of copying a tape. Let's face it, VHS has problems as-is without adding detrimental methods to the process.
Panasonic PV-8661 VCR via composite, then Toshiba RD-K2 DVD recorder used as pass-thru line-level TBC and s-video converter, then ATI AIW 9600XT AGP capture card via VirtualDub to huffyuv AVI.

[lordsmurf]

Better VCR may help that tape, preferably one with sharpening options.
AG-1980 would be my choice on this tape.