Video in PAL and Audio from NTSC source

[jairovital]

Hi there

I have one movie burned in two diferent DVDs and diferent ways. First one is in PAL version and the other is in NTSC. I've noticed that the PAL video was better (visual quality) than the other, in NTSC. PAL version was dubbed with portuguese audio, but the other one, NTSC was in english and original sound. So I thought to make a DVD using the best video (PAL) and the original sound (NTSC).

First I ripped both to my HD using DVDDecrypter. Using DVD-Lab I striped out VOBs to elementary streams (MPV and AC3). I put together that MPV video (PAL) and the AC3 audio (from NTSC source) and expected for result. As you already knew audio and video synchronism had gone... They started well, I agree, but they become far from each other along the time... That it because the PAL framerate of 25 fps and NTSC framerate of 29.97 fps, I thought

So, I used pulldown 2:3 to change 25 fps to 29.97 fps. Time duration of PAL video is 2h:03min:10s and NTSC is 2h:09min:20s. When pulldown had finished its work the video result is same time duration os NTSC video. Then I said, Bingo, it works! But now I've got another problem... DVD-Lab is refusing to compile the result video claiming the vertical frame size. In fact, my project is PAL, with 720x576 resolution, but video framerate is 29.97. Exact words of DVD-Lab is "Wrong video vertical frame size". If I change resolution of the video or change project properties, the lost of correct proportion may occur...

So what? Did I make the right thing? Pulldown is the correct answer for that problem? Is there another way to do the job? Is there a workaround to force DVD-Lab to do his duty?

Thank you all guru guys

[manono]

Do its duty? Your M2V is non compliant. DVD specs won't allow 720x576 and 29.97fps. Therefore it can't be authored. Easier might be to speed up the audio instead, to match the original video length. Or reencode the video to 720x480 to match the audio's length and make an NTSC DVD. You could try DVD Patcher to patch the headers to PAL 25fps and maybe DVDLab Pro will then think it's a true PAL M2V and allow it to be authored. I don't know for sure if that'll work because I don't believe in such tricks.

So, I used pulldown 2:3 to change 25 fps to 29.97 fps. Time duration of PAL video is 2h:03min:10s and NTSC is 2h:09min:20s. When pulldown had finished its work the video result is same time duration os NTSC video.

I don't believe you. If the original video is 25fps, applying pulldown of 25->29.97fps doesn't change its length. Now, if you apply pulldown set for 23.976->29.97, then it'll become the same length as the NTSC source.

[jairovital]

Thank you manono. I've read your reply several times to get the point. Of course you are right saying that my DVD specs is non compliance. I knew that, so I asked for help to cheat DVD-Lab.

The good news are your suggestion did worked! I will do more fine tunning to syncronize video and audio later, but I think it worked. I used DVD Patcher and change 576 to 480 therefore patcher was saying that it would go to unpredicable results. But not. DVD-Lab now accept it without claiming and did compilation process. Movie now has the same duration of NTSC one, despite it is 720x576 resolution.

I said before I need to syncronize video and audio later because PAL movie has some other presentation before and I must add some silence sound in the AC3 file to match NTSC time. But I gess it will match. If not, I will look for your advices again, if not disturbing you.

Once again, thank you manono

[Anonymous4]

12

[jagabo]

I don't believe you. If the original video is 25fps, applying pulldown of 25->29.97fps doesn't change its length. Now, if you apply pulldown set for 23.976->29.97, then it'll become the same length as the NTSC source.

This kind of blows my mind. I mean, I think I understand what you mean of course. Changing the frame rate doesn't change the length using a 2:3 or 3:2 because you are not just taking the max frames of the movie, calling a second every 25 or 30 frames. Instead you are manipulating the frames within a 1 second to correspond with a new frame rate.

Think of it like this: you have 24 frames and you need to display then 30 times in one second. I can tell you to display every 4th frame twice. Now when you play those 30 frames at 30 fps they will play for exactly 1 second, just like the original 24 frames played at 24 fps. 3:2 pulldown does exactly this -- but it does it on a field basis, not a frame basis.

When I do a frame by frame on this PAL mpeg, some frames look like half of the previous frame and half of the following frame. I'm guessing this is what happens when you go from 29.97 to 25 as is the case with this movie. (Japanese movie released for Italian market)

We'd have to see a sample. This could be field blending or just interlace artifacts.

What is the preferred solution to the aspect ration problem? Just trim a bit of the top and a bit of the bottom so that the 720x576 becomes 720x480 or compression the 576 to become 480?

No, resize the frame. If you crop the top and bottom you will have the wrong display aspect ratio.

[Anonymous4]

12

[jagabo]

I understand the 3:2 pulldown based on your explanation! Thank you. Good call on the field vs. frame. Is there a similar technique when going from PAL (25) to NTSC (29.97)?

Yes, 3:2:3:2:2. That gets you from 25 to 30. To get to 29.97 fps you have to change one of the 3s to a 2 every 1001 fields. Or drop a frame every 1001 frames.

[manono]

So I will not trim the video frame, instead I will resize it.

I'm betting this is a field-blended NTSC2PAL source. As such it's interlaced and you can't just blindly resize it. Nor can you apply pulldown to an interlaced source. You've already been asked once for a small sample of the source. Without such sample we're just guessing what you have and how to treat it.

[Anonymous4]

12

[Anonymous4]

12

[Anonymous4]

12

[manono]

Yes, it's field blended from an NTSC source. You sure you don't want to get the NTSC DVD of whatever this is? This one's pretty ugly and the only decent way to unblend it and restore it to progressive is via an AviSynth unblending script. Both of your 'chain of conversions' are lousy so I'm betting you don't know any AviSynth and thus can't improve it. My advice is to replace the audio without doing anything to the video. About the only good thing about this kind of NTSC to PAL conversion is that the length stays the same as the NTSC length, and the audio should fit OK without being reencoded. If you're lucky, at most you'll have to add in a delay. To replace the audio, follow Baldrick's Method 2 of his subtitle replacing guide. Replacing audio follows the same method as replacing subtitles in a DVD:

https://forum.videohelp.com/threads/277950-How-to-add-new-subtitles-to-an-existing-DVD

And then, if you don't have one already, get a DVD player that plays PAL DVDs.

[Anonymous4]

12

[AlanHK]

I WISH they had an NTSC release of this. I traveled to Japan and asked about it. No NTSC release (I went to Japan for other reasons but it never hurts to ask a few questions!). Disney would never allow an American release of this. The only Aladdin is their Aladdin apparently.

That's ridiculous. Aladdin doesn't belong to Disney. It's a thousand years old.

Anyway, there are lots of Aladdin movies and cartoons. (Same for Snow White, as some seem to imagine Disney created that story too.)

IMDB lists 68 titles, most of them non-Disney.

I think I will take your advice and just replace the audio. The concern will be that the taped video has commercials in it so I may just have include a bunch of delays. Before that I'm going to seriously need to reduce what I think is buzzing and some hiss noise. Back to digitalfaq!

I'd use Audacity (latest beta) for this.

Load the original soundtrack to use as a timescale, then add the English one as a second pair (presuming stereo) of tracks.
You might try its noise reduction filters, which are quite effective.

Then cut and slide the segments of the new track to match the original soundtrack. You can stretch or squeeze to fit if necessary. You might copy bits of the original track if there are missing parts, especially non-vocal.
Finally mute the original track and export your new version as AC3.

[jagabo]

If that MPEG does look ugly, its still the best that I've got to work with. I've had "need to learn AviSynth' on the back burner for a while so now is a good time to start I guess.

Wanting to learn something new is admirable, of course. But this video is probably not a good place to start. It has such severe problems (field blending, screwed up chroma channels) that you won't know whether the results you get are your fault or the video's.

[Anonymous4]

12

[AlanHK]

It seems that blended frames more or less alternate with unblended ones.

You can remove most (not all) of the blends by

DirectShowSource("aladdin.mpg")
Srestore(frate=25.0/2)

Then you can reinterpolate to 25fps if you want to make a new PAL DVD.

[Anonymous4]

12

[Anonymous4]

12

[txporter]

Wanting to learn something new is admirable, of course. But this video is probably not a good place to start. It has such severe problems (field blending, screwed up chroma channels) that you won't know whether the results you get are your fault or the video's.

Sorry to take this off-topic, but how does one go about determining that the chroma channels are screwed up? I am getting more and more comfortable with avisynth scripts but the luma and chroma stuff still elude me.

[Anonymous4]

12

[manono]

Sorry to take this off-topic, but how does one go about determining that the chroma channels are screwed up?

Unblend it and have a look. I wouldn't use AlanHK's script though, as the framerate after unblending is meant to be 23.976/24fps. And you'll want to bob it before unblending it. The default SRestore setting is OK. Here's a sample frame:

[Anonymous4]

12

[manono]

I just unblended it, cropped and resized it, using an AviSynth script, and played the script in MPC-HC. I went to a suitable frame and took a picture. I never actually reencoded it to anything. However, encoders, well most of them anyway, open AviSynth scripts just as they do video files. So, using these scripts you can reencode to any format you might like. Here's the script:

Yadif(Mode=1)
#TempGaussMC_beta1()#for use if you have all the time in the world
#TempGaussmc_Beta1u(1,1,0,0,0,0,edimode="Nnedi2",S Vthin=0.0,pelsearch=1)#ditto
Srestore()
Crop(14,2,-8,-2)
LanczosResize(480,360)

[Anonymous4]

12

[jagabo]

Sorry to take this off-topic, but how does one go about determining that the chroma channels are screwed up?

Unblend it and have a look.

You don't even have to go that far just to see the messed up chroma channels. A simply Bob() will show that the chroma channels are screwed up. You'll see something like manono's picture. I think only the bottom field of that video has the chroma channels screwed up -- it's chroma channels are a copy of the top fields chroma channels.

[Anonymous4]

12

[jagabo]

The top and bottom fields in the frame from which manono's picture comes are from two different film frames. Each field should have the chroma channels from its respective film frame. But the bottom field somehow got a copy of the top fields chroma channels. So its colors are from the wrong film frame. Ie, if the luma channel of the bottom field is from frame N, the chroma channels are from frame N-1.

film frame (N-1):


bad field (luma from N, chroma from N-1):


film frame (N):

[Anonymous4]

12

[txporter]

Thanks, manono and jagabo. I hadn't downloaded the video to take a look at it. I think even I could have spotted that one. How does that happen that the chroma channel get pulled from the wrong field?