dvd encode problems

[Morgoth]

I am trying to encode a dvd from a vhs tape which I have captured as an avi. I am framserving it from vdub to tmpgenc plus, and am using default pal dvd profiles. The end result is good quality, although with the noise filters at high quality it took 35 hours. The result did have one or two jitters in it which were not in the source avi though. Does anyone have any ideas what could do that, and how to stop that?

Also, this was done using pcm audio not mp2. whenever I tried to use mp2, I got an audio track that was out of synch, and the sound was about an octave out! Has anyone ever seen such a thing?

thanks,

morgoth

[Brian Ellis]

VHS tapes are not good quality, to start with and anything you do to "improve" the quality is a compromise. For example, noise removal looks for isolated ephemeral pixels and replaces them with surrounding colours. However, you can get apparent noise from the wanted signal as well, so that will also be attenuated. Horses for courses, 35 hours is horrendous for little gained and some lost. When "ripping" VHS tapes, I hardly ever bother with any kind of filtering, unless a shot really needs it. For example, I took a number of moonlit shots of wild animals in Africa and Asia on VHS/C. These, I adjusted in brilliance and contrast (and was surprised to find the colours coming up, as well), and the noise level increased, as well, as expected, mainly as blue/red speckling, a characteristic of PAL. I applied noise filtering to attenuate these, but heavy filtering also attenuated the stripes on a mongoose, for example. It was a compromise between noise and detail, although this was an extreme example.

PCM is standard for PAL DVDs.

[Morgoth]

thanks for the advice - I think I will just copy it over as is - only takes 8 hours on cbr.

Thanks for the info on the sound issue - at least I know now WHY it doesn't work!

morgoth

[Brian Ellis]

8 hours sounds very long. How long are your tapes and what are your hardware specs? I did 1.5 hr tape in just over 2h 12 min the other day at VBR 5800.

[Morgoth]

about 85 mins. My machine is a 256mb athlon 1333 on a very old amd irongate board - bad performance. Plus all drives are 5400rpm, so slow io. Perhaps thats why?

[Brian Ellis]

Yup!, it sounds like the CPU speed is the main cause. HDD speed is not really material for rendering, only capturing. The other slowing factor is that your memory is probably PC-100 with a slow FSB.

[Morgoth]

I think the memory is ok, - its ddr (2100), and I am planning on putting an xp2200 in soon. Hopefully that will improve it.

Curiously enough, the jitter problem is only on consumer units. I played the disk on software powerdvd and no jitter. Anyone got any ideas why it would be ok on software pc based dvd players and jitter a tiny bit in a few places on everything else?

thanks,

morgoth

[Peter B]

I would guess the jitter is due to interlacing. This is how I would do it (the best methods I have used by experimentation, information from here and other sources...)

First I deinterlace the video using AVIsynth and TomsMoComb both freely available.

Then open with virtual dub and add the levels filter, then apply dynamic noise reduction (free plug in). There are other filters that can be applied both in virtual dub and AVIsynth, but it is much easier to apply in Vitrual Dub as you can adjust them via a GUI as AVIsynth requires a script.

I highly recommend the deinterlace with TomsMoComb, I have used every filter I can find for deinterlacing via virtual dub and some are quite good but TomsMoComb is definately the best.

Then frame serve from virtual dub to TMPGenc, don't use any of it's filters, they are much slower. I usually use Virtual dub for cropping and resizing if required as well (faster).

Because you have deinterlaced it via AVIsynth set the source in TMPGenc to progressive and encoding to progressive (for some reason it always assumes the source is interlaced so make sure you change it to non-interlaced).

A lot of people recommend croma noise reduction for VHS and 2D cleaner and others. I would do a direct stream copy of a five minute segment of your footage (a bad part) and experiment with that until you happy and then run the full clip thorugh as above.

It is well worth mastering the AVIsynth TomsMoComb system

here is link with more details on how to do it and filters and when to apply them etc

http://arstechnica.com/guide/audio-visual/video-cleanup/cleaning-2.html

[Brian Ellis]

I agree that deinterlacing is better if you view on a computer but if the final output is for a TV set, then interlacing is very desirable. It sounds like you may already have deinterlaced as the pic is OK on PowerDVD. If so and you want it for TV, then recapture with interlacing and make sure you set the rendering and authoring programmes to the correct frame first.f

[Morgoth]

I believe it has captured interlaced, and the dvd authoring program part of tmpgenc has detected that. If interlacing was a problem, wouldn't the whole film jitter most of the time, not just 3 or 4 times over an 85 minute film?

[Peter B]

I always deinterlace if it is destined for DVD, the DVD player can interlace it if it is for TV as I find that interlaced mpegs have lower quality with more artifacts especially at low bit rates.

If you jitters are in fast moving areas then I think interlacing could be an issue. You don't tend to notice much with slow movement but when there is a lot and the differnce between fields is high it becomes a factor. (my opinion)

[Morgoth]

Well thanks for the advice regarding de interlacing all. I have completed a few test runs (long render times due to using tmpgenc on a slower machine), and once I de interlace, all works just fine - no jitters anyway when played on the dvd consumer units.

Now all i have to do is save for another processor.............


morgoth

[Morgoth]

Been having a few problems with de interlacing my mjpeg vhs video captures when converting them to mpeg2 for dvd.

I am framserving from vdub, and applying the vdub de interlace filter, and then running tmpgenc.

The resulting mpeg2 exhibits odd 'shadowing' effects - only noticeable on fast moving objects. Pausing and using frame advance shows a second 'shadow' edge behind the object with respect to its direction of movement. This effect is not present when I do not deinterlace the videos.

I have tried using the vdub filter with the blend setting, and changing the frame source in tmpgenc between interlace and non interlace makes no difference - the picture still 'shadows'. Could this be a problem with the field order(which I believe for a capture from a vhs pal source using avi_io and hardware mjpeg is bottom first?)or is there something more fundamental that I am missing?

thanks,

[Peter B]

I HIGHLY recommend using AVIsynth to deinterlace, however, if you want to use virtual dub, get the smart deinterlace plug in filter, it's much better than the internal filter.
go here for instructions on how to use AVIsynth and toms mocomb

http://arstechnica.com/guide/audio-visual/video-cleanup/cleaning-2.html

Look here for smart deinterlace

http://sauron.mordor.net/dgraft/smart.html

Frame only differncing usually works very well

[Morgoth]

It seems I cannot use avisynth and tomsmocomp - I do not have yuv source - it was captured via the hardware mjpeg, and avisynth complains when I try and load it in.

I tried the smart deinterlace for vdub but frankly it was just as bad. I have also tried the deinterlace in tmpgenc, all to no avail.

If I don't deinterlace at all, the picture is perfect on the TV, but am I likely to get the occasional jitter in the picture I saw before, which I refrerred to at the beginning of this thread? I do not care for deinterlacing particularly other than to sort out the jitter problem I had in one or two places on a capture I was working on. I have subsequently been deinterlacing based on the advice in this thread, and everything has been fine, but I have found lately on two other things I have attempted to capture and convert this shadow problem with the deinterlacing. Why has this not happened on all my captures - they are all off commercial vhs tapes or recorded vhs from satellite TV. Is this due to different methods of frame conversion?

thanks,

morgoth