Back to the PAL to NTSC conversion

[iGorland]

I have a 15 minute video of New York that I shot in December 1999 with the WTC still there. It is in PAL and I wanted to put it together with the new NTSC video that I will be shooting in NYC soon on the same DVD. However, it is not allowed to have two different systems on one DVD project, so I need to convert my PAL video to NTSC. I used VirtualDub to 1) deinterlace; 2) change the frame size; 3) change the frame rate. Then I encoded it by CCE. The resulting video was OK, except for some jittering in high motion scenes.

Now, it occured to me what if I record the video from my PAL camcorder to my NTSC camcorder through the iLink cable!!! Please, let me know if anyone tried this! Cheers.

[Xesdeeni]

I don't know how the error will show up, but you won't accomplish what you want that way.

Give my conversion script a try: http://forum.doom9.org/showthread.php?threadid=35387

Xesdeeni

[iGorland]

Xesdeeni. Thanks! In fact, I tried SmoothDeinterlace, and it seemded to me that it was no better than deinterlacing by VirtualDub. As I said, the result that i originally had was not bad at all, just some very small jittering. Well, using SmoothDeinterlace does not eliminate same insignificant jittering in high motion scenes.
Is it the case with you as well?
Cheers

[Xesdeeni]

Did you try SmoothDeinterlace with the AVISynth script (http://forum.doom9.org/showthread.php?s=&threadid=35387)? Please note that you cannot do the same thing with VirtualDub (alone). VD will not allow you to get the double framerate in the intermediate steps needed to minimize the jutter.

No, the results with the plugin and AVISynth are MUCH better. I would not have spent the time porting the plugin from VirtualDub, working with the AVISynth team to fix a bug, and creating the script if the results weren't much better than I could get with what was available (in VirtualDub or AVISynth).

The results are not as good as the most expensive of hardware conversion devices, which use motion estimation/compensation to estimate the position of objects for destination frames that actually fall between the time of the source frames. But the results are about the same as the less expensive conversion equipment (called four-field conversion), and can sometimes look better (IMHO).

But this technique will always have some type of jutter. If you use ConversionType=0, then the jutter will show up as two successive NTSC fields that were taken from the same PAL field (using VirtualDub, you'd have four fields from the same PAL field!). When film is converted to NTSC (or when your NTSC DVD player plays a film DVD), a 3:2 pulldown is done, which does something similar with field replication. I have found that to my (I have references that will vouch that I have the most critical eye around...seriously!) this is not any more noticable than the 3:2 pulldown.

If you use ConversionType=1, you will get something very close to the four-field (less expensive) conversion technique. In this case, there will be a jutter on pans caused by multiple PAL fields being included in the same NTSC field. I find this type of jutter extremely annoying on the PAL programs that are shown over-the-air here. Personally I prefer the above double-field to this technique, because it is much less noticable. [ I do think it might be preferable on NTSC->PAL conversions due to the complete loss of a source field, but without any PAL equipment at my disposal, I can only take an educated guess. ]

ConversionType=2 is yet another type of conversion that is basically an asynchronous conversion. That is if you pointed a (special) NTSC camera at a (special) PAL TV, you'd see parts of two PAL frames in each NTSC frame (the "special" part is to eliminate blanking and phosphor decay). [ The filter does allow some blending at the boundary to avoid tearing. ]

Xesdeeni

[iGorland]

Thanks, Xesdeeni! I will give it a try and will let you know. My best!

[morsa]

how did it work?

[iGorland]

It did not work. I cannot get my avisynth 2.06 to run on my computer. I installed the dll in winnt/system32, and then clicked on the reg file. However, the CCE read the resulting avs as 240 frames and 10 seconds long, which is not the case. The avs file is not readable by media player. I understand that I did not install it correctly, but do not understand what i did wrong. cheers

[Xesdeeni]

If your CCE read the file as anything at all, then you actually installed AVISynth correctly.

What you are probably seeing is a combination of things.

First, you can read .avs files in media player. But you have to do it backwards. Right-click on the .avs file in Explorer and choose "Open with..." In the dialog, choose the "Windows Media Player," NOT the Microsoft(R) Windows Media Player.

What you will probably see then is an error message. I think they are standard 24fps and 10 seconds long, hence your 240 frames.

You can also see the error message if you open the .avs file in VirtualDub.

Anyway, the error is probably in your script. If you are using the standards conversion script I wrote, then you probably have to make sure that the video is opened correctly.

Xesdeeeni

[iGorland]

Thanks, Xesdeeni! Looks like I am getting into Avisynth Forum issues now, so if you want, we may continue in that forum. May the problem be because I am trying to frameserve from Premiere using videoserver 0.93 plug-in? My best!

[Xesdeeni]

I have frameserved via AVISynth from Premiere, but I don't recall what I installed to make it work :rolleyes: , and it's been a while since I did this, so I don't recall how Premiere and AVISynth interacted (and I'm not sure whether AVISynth 2.x works).

I suggest you export the edited video from Premiere and make a temporary AVI file and use that as an input to the script. Then use VirtualDub (or something) to open the script and save yet another temporary AVI file that you can use in Premiere with your other PAL video. 15 minutes will fit in 4GB (if you are using Win9x). If you are just cut editing, then the first temporary file won't be any different than the direct output from Premiere (if you have text or fades, then it will have to be decompressed and then recompressed).

Xesdeeni