Problems capturing from VHS to MPEG.

[Deane Johnson]

I'm new, but trying to equip myself to do this right.

I tried capturing an LP VHS video to make a DVD of it. Using Pinnacle Studio 9.1, it looked fine on the small screen. When playing back the MPEG, I noticed severe vertical banding or streaks, especially in bright areas. They were fairly narrow.

I went ahead and made a DVD, the vertical bands were much reduced in intensity, but the entire picture had diamond shaped artifacts throughout, evenly spaced.

My equipment:

NEC S-VHS into a TBC-1000, then into a Cirrus Logic capture device (the one that comes with Studio 9). I used S-Video all the way through.

I used a trial copy TMPGEnc DVD Author to record the DVD +RW at x2.4.

Any idea what is causing my problem.

Thanks

Deane

[Poplar]

With SVHS and TBC, you should get quality very, very close to that of the original tape. Wish I had them.. Could you be more specific with the settings, such as bitrates? Anything under 4Mbps, artifacts won't be transparent...

[FulciLives]

You definately got good equipment with the S-VHS VCR and the TBC there.

However LP and SLP/EP speed VHS tapes don't look very good and require some filtering work ... even when played back with good hardware as you have.

I'm not familiar with your capture hardware (the capture card/device) but if you can capture to an AVI file (using PICVideo MJPEG or HuffyUV) then do that and then try a good "video noise" filter such as Convolution3D or the one built-in to TMGPEnc Plus etc.

Here is a link to a guide I wrote on using Convolution3D:

https://www.videohelp.com/forum/viewtopic.php?t=225951

That should help clean up the image a bit.

Also the worse the video quality the higher the bitrate you need. Often times with low quality sources like this you are better off using Half D1 resolution (352x480 NTSC or 352x576 PAL) and then MAX out the bitrate. For instance Half D1 will max out around 4000kbps to 5000kbps so if the running time is short enough to be in that range you can even just do a simple CBR encode.

If you can get up to 6000kbps due to a short running time then a 2-pass or multi-pass at Full D1 (720x480 NTSC or 720x576 PAL) should be fine. Full D1 will also look fine using CBR if you can get the bitrate up to 7000kbps or higher ... all depends on the running time of what you are encoding.

Chances are though that with a LP or SLP/EP speed VHS as a source that any resolution difference between Half D1 and Full D1 will not be noticeable.

Therefore you are better off going with the lower resolution of Half D1 because the final encode will have very little if any MPEG artifacts with the lack of reslution in Half D1 not even being an issue (due to the low quality and lack of resolution in the original source).

If you intend to use Half D1 resolution you are still (in most cases) better of doing your capture at Full D1 resolution then resize to Half D1 during the encoding stage. Also make sure you do your filtering BEFORE you resize down.

- John "FulciLives" Coleman

[Poplar]

Very interesting. Is Convolution 3D still the best denoise filter around? Haven't been to that front for quite a while....

[FulciLives]

Very interesting. Is Convolution 3D still the best denoise filter around? Haven't been to that front for quite a while....

I think it is still considered the best all around "video noise" filter.

The all mighty powerfull geek gods (I say that with affection since I have learned a lot from them) over at doom9 seem to favor using a varied combination of "lesser" filters to get a desired effect (based on the source) but they are a bit insane (plus I haven't really been there lately) but last time I was looking at the AviSynth forum there on doom9 they were doing some pretty "wacky" looking combinations of filters!

So for the regular non-diety mortal man I think Convolution3D is still king. Hell even the TMPGEnc Plus "video noise" filter is damn powerfull but I stopped using TMPGEnc Plus in favor of Cinema Craft Encoder due to colorspace (i.e., luminance) issues.

Since most captures are YUV you are better off leaving it that way from start to finish. This can be done if you use AviSynth and CCE but not so with TMPGEnc Plus ... nor is it good to use VirtualDub frameserving because again you are in an RGB environment and thus a colorspace conversion.

- John "FucliLives" Coleman

[piano632]

Sounds like you're getting some kind of electrical interference thru the cables. Do yourself a favor and get some good quality shielded cables. Most of the cheap cables that come in the box with equipment are so poorly constructed that they actually worsen picture quality.

Here's where I get most of my cables. These are about the best you can get at any price, and I think they are very reasonably priced considering how much improvement they make.
http://www.signalcable.com

You might want to also consider using a JVC standalone recorder - they do a great job of removing grain and chroma noise from video before converting to MPEG. Some people are saying JVC recorders make the best-looking VHS to DVD transfer quality.

[Deane Johnson]

Wow, some great answers and suggestions here. I appreciate them very much. Looks like I've got a few days of experimenting and testing ahead of me. I'll report any solutions.

Poplar, I think the bitrate was 6Mbps. Every available setting was set for maximum quality.

Deane

[lordsmurf]

Poplar, I think the bitrate was 6Mbps. Every available setting was set for maximum quality.
Deane

By whose assertion? This may be the problem right here.

[Deane Johnson]

Explain a bit further. I'm not as knowledgeable as the rest of you guys.

Perhaps I should explain, I used Pinnacle 9 to capture. There aren't a lot of settings. Sort of good, better, best choices. Is there something more controllable I should be using?

As mentioned in the opening post, this is all new to me. I don't even know what the abreviations mean much of the time. I go to bed with a headache at night after cramming information into my 70 year old brain. But, thanks to the great information on this forum, I am learning.

Thanks

Deane

[lordsmurf]

That's a horrible choice in capture software. I would capture AVI, encode MPEG-2, and then author with TDA (as you're doing already).

[gshelley61]

Yeah, my analog caps with Pinnacle 8 Deluxe were never very good. The software is buggy and slow, too.

Their MPEG2 encoder is a POS, as well. Macroblocks!

[piano632]

I am always mystified by the differences of opinion regarding Pinnacle Studio. Some people say how great the quality is and others say it's crap quality. I also have Studio 8 Deluxe and I have no complaints about it's capture quality or MPEG encoding quality. In fact I have compared Pinnacle's MPEG encoding to TMPGEnc and saw very little difference. I will admit that Pinnacle does tend to oversaturate colors (at least with my card), but that's easy enough to fix by turning down the saturation level during capture. I must wonder if these differences in opinion are because Pinnacle is using different chips in their capture cards. My Studio 8 card uses a LSI chip, and now I see by this thread that Studio 9 is using Cirrus Logic. Maybe I got lucky and got a good Pinnacle card, but I would be curious to know what chip everybody has in their card.

[gshelley61]

I got better capture results on my computer with VirtualVCR and an AverMedia card than I did with the Pinnacle hardware and software.

Then, when I tried out a used Panasonic DVD recorder on a whim, it was the the first time I felt like I achieved the video transfer quality I was looking for. The JVC makes an even better looking video recording (MPEG2 capture) than the Panasonic did...

I think the quality difference has to do with superior video input and sampling sections of DVD recorders (designed and built by major electronics firms with literally decades of experience in consumer and professional broadcast A/V products) versus capture cards made by computer software and hardware companies.

[Wilbert]

convolution 3d: I think it is still considered the best all around "video noise" filter.

It's quite useless (meaning there are filters around which are much better) for cleaning your analog caps.

For that case I recommend: PeachSmoother, RemoveDirt or TTempSmooth. The first can be found here:

http://www.avisynth.org/warpenterprises/

The latter two can be found here:

http://forum.doom9.org/showthread.php?s=&threadid=84481

and like lordsmurf says (free translation ) use huffyuv or mjpeg for capping.

[piano632]

You be the judge and tell me if you see anything wrong with these caps. All 3 were made from analog cable into JVC HR-S9600U as a tuner then into Pinnacle AVDV card, converted to MPEG2 @ 5500 bps using Pinnacle MPEG encoder, capped with PowerDVD and converted to JPEG @ 90% quality with IrfanView.

[Deane Johnson]

Update!

Good progress, but not there yet. After going to Best Buy, CompUSA and our local NFM, I found one AVI USB 2.0 capture device at CompUSA. It is a Dazzle. I was disappointed when I opened it to see Pinnacles name on it and another copy of Studio 9. Oh well.

After getting a driver installed, I had video. However, I had lots of trouble getting a capture going. Would you believe Studio 9 would recognize the Dazzle, but would recieve no video. And, that's the software packaged with it.

Sony Movie Studio (neat program) will only recognize MPEG. So, back to the trusty old Roxio 7 Creative Suite. It got a picture, but no audio. I went ahead an captured about 5 minutes of the LP VHS tape (wow, does that eat up hard drive), converted it to MPEG with TMPGEnc, authored it to a DVD+RW using DVD Lab (I like this program), then played it pack. Wow. At first blush I'm prepared to say it looked as good, or perhaps better than direct from the tape. The vertical banding was gone, the diamond shaped artifacts were gone. But no audio.

Why I need now is to figure out a good capture program that will let me monitor both audio and video during the capture. I need some basic controls like "color saturation". This particular tape is grossly over saturated and has to be cranked way down, but then the result is good once it's done.

Roxio allows this, but I need to solve the audio problem. Any ideas for software, or what my problem may be. I feel I'm on a roll with the help I'm getting here.

Thanks again.

Deane

PS: Piano632: Those are gorgeous captures. After I get the above problems solved, I'll move onto questions on using a DVD recorder for even better quality ( I assume, but that's another subject).

[piano632]

Why I need now is to figure out a good capture program that will let me monitor both audio and video during the capture. I need some basic controls like "color saturation". This particular tape is grossly over saturated and has to be cranked way down, but then the result is good once it's done.

Pinnacle Studio has controls for color saturation, brightness, contrast, sharpness and hue. You can find these controls to the left of the pie graph that shows hard drive space. Click on the TV icon to open up the video control panel. To the right of the pie graph is the speaker icon which opens the audio level control panel. (I have Studio 8 so the layout might possibly be different with Studio 9.)

P.S. You shouldn't be surprised to find Pinnacle products inside Dazzle boxes since Pinnacle bought out Dazzle some time ago. Also Pinnacle software is notorious for not recognizing any other brand of video card aside from their own. (Well actually it may recognize it, but it won't let you capture with it if it's not a Pinnacle card.)

[Deane Johnson]

I can't get a preview picture operating in Pinnacle Studio in order to use the color control. Also, I don't get any audio.

Meantime, I can get a picture in Roxio Creative Suite 7 and have full color controls. In fact lots of controls such as the ability to mask the flicker at the bottom of the frame on some tapes. But no audio on this program either.

This audio thing has me puzzled.

Deane

[FulciLives]

I can't get a preview picture operating in Pinnacle Studio in order to use the color control.

Be carefull with picture controls like brightness, contrast, color saturation etc. because NTSC looks different on a PC monitor than on a TV.

In short you can't trust your eyes when adjusting such picture properties on your computer monitor.

- John "FulciLives" Coleman

[bhutchins]

return the razzle dazzle and go get the Canopus ADVC-100
rated tops at this site, which I contest to. Most any program will recognize it.
https://www.videohelp.com/capturecards.php?CaptureCardRead=132#comments

[offline]

convolution 3d: I think it is still considered the best all around "video noise" filter.

It's quite useless (meaning there are filters around which are much better) for cleaning your analog caps.

For that case I recommend: PeachSmoother, RemoveDirt or TTempSmooth. The first can be found here:

Have to agree with that. I trashed Convolution 3d for peach
pretty quickly many many moons ago. Depending on your source, three or four filters
descretely applied is much better than one on set medium to high. Convolution
is more of a harsh "fix it all" for those that need a basic script.

[NamPla]

Since most captures are YUV you are better off leaving it that way from start to finish. This can be done if you use AviSynth and CCE but not so with TMPGEnc Plus

You can install the "ReadAVS" VFAPI plug-in and TMPGEnc will accept YUV Avisynth scripts.

I like TMPGEnc because you can crop the borders & centre the video neatly. This helps with overscan issues, and doesn't affect the aspect ratio (if you've captured at 704/resized to 352).

Also the "sharpener" filter seems (to me) to be very good, better than complicated Avisynth filters/functions/tweaking variables/etc... I just plug in "horizontal = +100" and "vertical = -100", and the sharpening effect (after Convolution3D) is GREAT (ie no artifacts)!

Maybe CCE is better, but better the devil you know... :P

[Deane Johnson]

return the razzle dazzle and go get the Canopus ADVC-100
rated tops at this site, which I contest to. Most any program will recognize it.
https://www.videohelp.com/capturecards.php?CaptureCardRead=132#comments

Pardon the elementary question, but can one capture AVI with this?

Deane

[FulciLives]

Since most captures are YUV you are better off leaving it that way from start to finish. This can be done if you use AviSynth and CCE but not so with TMPGEnc Plus

You can install the "ReadAVS" VFAPI plug-in and TMPGEnc will accept YUV Avisynth scripts.

I like TMPGEnc because you can crop the borders & centre the video neatly. This helps with overscan issues, and doesn't affect the aspect ratio (if you've captured at 704/resized to 352).

Also the "sharpener" filter seems (to me) to be very good, better than complicated Avisynth filters/functions/tweaking variables/etc... I just plug in "horizontal = +100" and "vertical = -100", and the sharpening effect (after Convolution3D) is GREAT (ie no artifacts)!

Maybe CCE is better, but better the devil you know... :P

TMPGEnc Plus is an RGB environment as is VirtualDub.

So if your capture is YUV (which most are) then you need to do a colorspace conversion. This can lead to luminance level problems.

- John "FulciLives" Coleman

[NamPla]

Yes but I read this over at the kvcd.net forum:

It's important to note that you must be outputing in YUY2 format from Avisynth for the plugin to work. Never use the "ConvertToRGB" Avisynth filter at the end of your script, or it will not load into TMPGEnc.

So what exactly is this plug-in doing? Is it by-passing the RGB conversion or what???

I also want to query some people's criticism of Convolution3D. Are you people capturing VHS? Because that's all I am doing, and Convo3D (in particular, the "animeBQ" preset) does wonders. I've briefly looked at things like PixieDust & PeachSmoother & even MipSmooth, but they just don't do the job as good. If I start adding a string of filters to my script, my encoding time increases. I simply cannot give a bad opinion of Convolution3D (especially when using the TMPGEnc "sharpener"! ( - which adds next-to-nothing to encoding times!)).

Maybe if I had "pro" S-VCRs & cables & A/D bridges etc, maybe then I might consider Convolution3D differently...? Who knows. Each to their own I guess.

Cheers.

[Wilbert]

Are you people capturing VHS? Because that's all I am doing, and Convo3D (in particular, the "animeBQ" preset) does wonders. I've briefly looked at things like PixieDust & PeachSmoother & even MipSmooth, but they just don't do the job as good.

Yes, I cap analog stuff. I didn't mention Dust, because sometimes it cause artefacts. Imo, animeBQ blurs very much.

If you want to use Peach, read the documentation, and start with

PeachSmoother(clip, dot=true, readout=true)

Play the video (frame by frame), when the green dot appears write down the NoiseLevel and Baseline values, and use them

PeachSmoother(clip, noiselevel=..., baseline=...)

MipSmooth also blurs too much when you use it as a spatial filter ....

Did you try RemoveDirt and TTempSmooth?

[Deane Johnson]

As the starter of this thread, I'll update my situation.

I appreciate the excellent discussion which has been very helpful.

I've decided to quit messing around with this Pinnacle junk and go up a couple of levels.

I found the Pinnacle forum and discovered that many people were having comparable problems with the software, drivers, lack of program recognition, etc. And very little help from Pinnacle.

I took the advice of bhutchins who posted earlier in this thread and checked into the Canopus ADVC-100. The Canopus guys were very helpful on the phone and as a result I have one being overnighted, so I'll have it tomorrow. Hopefully this will be the end of my problems.

The cheap stuff always ends up costing more in the long run because you have to replace it with the good stuff at some point. Unfortunately, I had little knowledge at the outset, so walked into the Pinnacle trap.

Thanks for your help.

Deane

[NamPla]

I hope no body minds me ressurecting this thread... It's interesting & should be discussed more, I think.

Wilbert's recommendations certainly got me testing away again. As a self-confessed "noob", I realize now I was testing all wrong before. (as I'm dealing with interlaced sources, etc, I had to get my scripts right).

However: I am sticking with Convolution3D.

I have tried (and had great success) with things like "Deen/Undot" and "Remove Grain/Remove Dirt". I don't see much else that comes close to these, except for Pixie Dust, which is slow as hell. MipSmooth is OK. I guess it all depends on your hardware, and your taste.

But I find this script the best:

LoadPlugin("C:\~\Convolution3D.dll")
LoadPlugin("C:\~\Undot.dll")
Import("C:\~\InterlacedFunctions.avsi")
avisource("C:\~\capture.avi",false)
UnfoldFieldsVertical(true)
Convolution3D(preset="animeLQ").Undot()
Lanczos4Resize(352,576)
FoldFieldsVertical(true)
ConvertToYUY2(interlaced=true)

The reason I stick with Convo3D (despite everyone's claims to the contrary) is that it is sharper than the others. Firstly, the "Remove" filters are a bit more blurry. They're great, don't get me wrong, but just a bit blurrier. And "deen/undot" combo is great, but requires a YV12 colour conversion. And Peach is good too, but I had artifacts with it, and researching around I discover other people have too. It's not too good with sudden movement.

Convolution3D does the job! I confess I have "progressed" from preset = VHS to AnimeBQ now to AnimeLQ. AnimeLQ seems to do it for me. You could even go to MovieLQ I guess...

I throw Undot in, because it doesn't add to encoding time, but helps out a bit.

And I have tried lots of different sharpeners. I don't like them any more. I don't use them. They cause artifacts & increase encoding time. I have tried Asharp & LimitedSharpen etc, and while it looks good in VirtualDubMod, after encoding you can definitely see them working. I don't like it. You gotta remember, when playing on the TV, everything looks better anyway...

Just my opinion, but I thought I'd throw it in for discussion...!?

Cheers[/quote]

[SatStorm]

I'm old fashion Virtualdub guy...
I capture 768 x 576 (picVideo ( 19 )) using virtualdub mpeg 2. Then I filter:
Sharpen ( 6 ) - Resize ( 352 x 576 ) - rmPAL (for PAL captures) - Static Noise Reduction ( 6 - Interlace ) - Dynamic Noise Reduction ( 8 )

Frameserve to TMPGenc plus 2.5. Encode 352 x 576 CQ 2000-5000@65%

I'm pleased with the results, about 3.30 - 4.15 min per DVD5 and realtime encoding on my AMD 2600XP

Of course, I'm PAL. With NTSC that YUV > RGB > YUV could be a problem....

[NamPla]

SatStorm, how do you get realtime encodes? I have knocked it down to 2x-3x, but...? Maybe picVideo is faster than huffyuv (what I use)?

Also doesn't 768 -to- 352 screw up the aspect ratio? (I'm not sure, that's the idea I got from reading around...). I was under the assumption you had to cap at 704 if you were gonna downsize to 352?

Cheers