3h home movie VHS to single DVD without visible artifacts?

[atl]

Ok, I've been trying this for a few weeks now, with increasingly time
consuming methods (started with Ulead VS6 added TMPGENC noise reduction, then VBR, high Q motion search etc).
Is it possible?

I capture at high bitrate MPEG (10kbps, I-frame only), and all looks fine.
But all my attempts to squeeze a 3h VHS tape onto a single DVD (i.e. about 3.2 kbps) result in either motion arctifacts or loss of resolution/ macroblocks.. Things are getting better but not good enough!

If anyone has achieved a 3h DVD from home movie VHS and can't tell
the difference with the original (or improved it..) I 'd like to know how they did it.

[TeeeRex]

Have you tried to convert the captured file to 1/2 DVD resolution instead of the full resolution - on an original VHS source I doubt you will notice this - and of course you then then double the bitrate to 6.4 kbps which should be suitable for most things.

Other thing - I noticed that you say you are capturing straight to MPEG - is your hardware card doing this ? - If so, then do you have the option to capture uncompressed (or lossless compression) to AVI which will give you a much better source to start with ? Try this with a small sample first as disk space will be your main problem here ...

Good Luck ...

[jimmalenko]

Have you tried to convert the captured file to 1/2 DVD resolution instead of the full resolution - on an original VHS source I doubt you will notice this - and of course you then then double the bitrate to 6.4 kbps which should be suitable for most things.

errrr, no ?

Halving the screen size does not mean you can double the bitrate - the bitrate will need to be exactly the same (3.2). It will look a helluva lot better at that bitrate, though.

[lgh529]

The problem with capturing home movies from VHS tapes is that the video noise reaks havock on the MPEG encoder regardless of the engine. You could probably get close to three hours on a DVD at full resolution if you use a decent TBC, using a deshaker filter in VirtualDub, and then encoding with VBR.

When I need the best quality, I use all the noise filters that I can make work, which is going to increase your time significantly. For example, I just converted my old VHS tape of Apollo 13 using my JVC HR-S9911 SVHS VCR. Because it is professional quality with little noise and no need for deshaker it only took 4 hours to encode. The length is about 2 hours 20 minutes. With 224Kpbs AC3 audio, my DVD only took 3.3GB.

The last wedding video I did however was a different story. The VHS tape was made in 1985, almost 20 years old. The tape was in terrible condition and the camera person was lousy. After using deshaker, color correction, noise filters, etc., I encoded 51 minutes of video that took 22 hours.

Moral of the story: Home videos take a lot longer to do, and when you are looking for 3 hours worth, your noise tolerance must be very low.

[Capmaster]

I capture at high bitrate MPEG (10kbps, I-frame only), and all looks fine.
But all my attempts to squeeze a 3h VHS tape onto a single DVD (i.e. about 3.2 kbps)...

and of course you then then double the bitrate to 6.4 kbps which should be suitable for most things.

Better get the units straight or some noob will really try 6.4 Kbps and wonder why his video looks like it was drawn with crayons It's "Mbps" - megabits per second, not kilobits per second. Alternately you can write it as 6400 kilobits per second, or 0.0000064 Terabits per second for that matter, but the correct unit is megabits per second. Also, "bits" is written with a small "b" and "bytes" is written with a capital "B".

I don't mean to be preachy, but there is a big difference between "K" and "M"

[BSR]

I agree with all the above.
Capture uncompressed or some other less lossey codec than MPEG2.
Use filters.
Encode to MPEG2 with TMPGenc with Half D1 resolution (352x480)

If you want Full D1 (720X480) put less video on the DVD. No more than 2 hours.

Or for a test try captureing at MPEG2 at Half D1(352x480) and see how your results come out. It will save a bunch of time, but quality will suffer slightly(all depends).

[FulciLives]

Capture Full D1 (720x480/576 NTSC/PAL) then when you re-encode the capture use Half D1 (352x480/576 NTSC/PAL) and use some sort of "video noise" filter.

One of the best "video noise" filters is Convolution3D for AviSynth. Try the setting called "vhsBQ"

As for the bitrate ...

To be safe I would use 3100kbps because 3200kbps might be pushing things a bit too close to the edge of the disc. Of course I use 256kbps AC-3 audio so 3200kbps might be OK if you are using 192kbps AC-3 audio. I would never go lower than 192kbps for AC-3 and if you are using MP2 I wouldn't go lower than 224kbps.

Also make sure you do a 2-pass VBR. I would try 1000kbps MIN, 3100kbps AVG and a MAX of either 5000kbps or 6000kbps

The real key here to getting the best is the filter though. Even Half D1 at those bitrates can look bad if the source is not filtered UNLESS that source is damn nice to begin with (such as a DVD rip).

- John "FulciLives" Coleman

[TeeeRex]

doh!

That'll be me all embarrased then ...

[atl]

Thanks all for the suggestions..

Yes I do hardware MPEG capture at 9.9 Mbps (not kbps..). Saves me lot of diskspace compared to avi and I can't tell the difference with the source
if I burn it straight to DVD, so my problem is really a compression problem (9.9 -> 3.1 Mbps)

Looks like better noise filtering is going to be the answer.
I'll definitely give the Convultion3D/vhsBQ filter a go, once I figure out how to feed my mpeg source into avisynth into TMPGEnc.

If anyone else has specific suggestions of filters (and details of settings)
please let me know.

[deckard8]

I take it you mean 'convolution3d' and not 'convultion3d'. Although after enduring a 2 day wait while your video is encoded and filtered may well induce 'convultions' in spectacular 3d! :P [/u]

[atl]

Yes, my writing was as convoluted as the process to get to a respectable result.

If anyone has suggestions on how to speed up the interminable filtering and encoding process I would be very interested, anything longer than overnight is starting to get a bit much..
Is it just CPU speed that counts? (I've got a 2.6GHz/400 MHz FSB/HT machine, so not much more than 25% to be had from upgrading to the latest and greatest. Suppose extra cache might help)

[Capmaster]

Yes, my writing was as convoluted as the process to get to a respectable result.

If anyone has suggestions on how to speed up the interminable filtering and encoding process I would be very interested, anything longer than overnight is starting to get a bit much..
Is it just CPU speed that counts? (I've got a 2.6GHz/400 MHz FSB/HT machine, so not much more than 25% to be had from upgrading to the latest and greatest. Suppose extra cache might help)

Probably as big a factor (up to a point) as CPU speed is HD speed. Get yourself as fast a HD as you can if you plan to be doing a fair amount of video work. Go for the read/write transfer rate, as most video work results in huge files being moved around. Gaming is known for taxing the CPU and the graphics system to the limits ...video editing taxes the HD to the same extent. CPU speed helps, but HD speed will make the biggest difference.

[qlizard]

If you capture straight to mpeg, you're not going to be able to filter it without re-encoding, causing further quality loss. One of the disadvantages of capping straight to mpeg. Unless someone knows a trick I haven't heard of.

[Capmaster]

If you capture straight to mpeg, you're not going to be able to filter it without re-encoding, causing further quality loss. One of the disadvantages of capping straight to mpeg. Unless someone knows a trick I haven't heard of.

Oh, I don't know about that. My Snazzi III USB2 (analog to MPEG) has a whole slew of filters and controls available for chrominance, luminance, VHS noise, sharpness, AGC, white peak, luminance phase, H & V input window start, length and end, luminance scaling increment, color balance, brightness, contrast, saturation, raw gain control, audio/video clock lock, many others. Of course they're not mentioned in the manual - I had to get the datasheet for the Philips SAA7114 Video Decoder/scaler which it uses for the front-end. All the features are available via registry settings.

The Dazzle DVCII is another capture device that allows this sort of control if you know what to change

[SLK001]

I once put four 50 minute episodes of a TV series captured from VHS to DVD. I used FULL D1 resolution (720x480). The tapes were old and had considerable noise on them. I used CCE with the NOISE FILTER set rather high (IIFC, 20 out of 25). The resulting videos had CCE defined "Q" levels under 5 ACROSS THE BOARD (for those familiar with CCE, there were no "GREY" areas in the ADVANCED SCREEN when viewing the "quantization" factor).

Three hours of video on one DVD should be a piece of cake.

[FulciLives]

Run the captured file through DVD2AVI. I am assuming your captured MPEG-2 file is an MPEG file i.e., filename.mpg

So run that through DVD2AVI and this will split out/spit out your audio file (probably MPA/MP2 format) and give you a D2V project file.

You can then READ the video into your AviSynth script using the D2V project file.

Your script would look something like this:

LoadPlugin("mpeg2dec3.dll")
LoadPlugin("Convolution3d.dll")
mpeg2source("D:\filename.d2v")
ConvertToYUY2(interlaced=true)
SeparateFields()
odd=SelectOdd.Convolution3D (0, 32, 128, 16, 64, 10, 0)
evn=SelectEven.Convolution3D (0, 32, 128, 16, 64, 10, 0)
Interleave(evn,odd)
Weave()

This script works AS IS with Cinema Craft Encoder. If you are using TMPGEnc then add this as the last line:
ConvertToRGB(interlaced=true)

All that SeparateFields() and Weave() stuff is needed since we have an INTERLACED video file as are all captures (unless your card deinterlaces which is BAD if it does).

- John "FulciLives" Coleman

[qlizard]

My Snazzi III USB2 (analog to MPEG) has a whole slew of filters and controls available for chrominance, luminance, VHS noise, sharpness, AGC, white peak, luminance phase, H & V input window start, length and end, luminance scaling increment, color balance, brightness, contrast, saturation, raw gain control, audio/video clock lock, many others. Of course they're not mentioned in the manual - I had to get the datasheet for the Philips SAA7114 Video Decoder/scaler which it uses for the front-end. All the features are available via registry settings.

The Dazzle DVCII is another capture device that allows this sort of control if you know what to change

yeah, I was mainly referring to filtering after the capture. My Dazzle 150 also has some settings for adjustment during the capture.

[Capmaster]

My Snazzi III USB2 (analog to MPEG) has a whole slew of filters and controls available for chrominance, luminance, VHS noise, sharpness, AGC, white peak, luminance phase, H & V input window start, length and end, luminance scaling increment, color balance, brightness, contrast, saturation, raw gain control, audio/video clock lock, many others. Of course they're not mentioned in the manual - I had to get the datasheet for the Philips SAA7114 Video Decoder/scaler which it uses for the front-end. All the features are available via registry settings.

The Dazzle DVCII is another capture device that allows this sort of control if you know what to change

yeah, I was mainly referring to filtering after the capture. My Dazzle 150 also has some settings for adjustment during the capture.

Yeah, it would be nicer to be able to filter after the capture, but I usually don't need to do much filtering except for the VHS noise. The Movie Mill app I use has a real nice filter that results in a VHS capture that is better than the source. Too much filtering and everything's soft. Not enough and you get noise specks that eat up the bitrate and don't look good. MM seems to have just the right level of filtering.