DVD Pal to DVD NTSC

[DJRumpy]

Certainly you can. You would need to deinterlace your PAL source, and then slow it down by 1 frame per second. The output would have deinterlace artifacts, but the conversion process isn't that difficult.

[Bluesilo]

ok i read xdeen.... site again and saw i needed to go to 29 fps but now im getting the interlace line all across my screen when I play back im not sure what to fix in tmpgenc settings but im using the same script as seen above only

CovertFPS is 59.94

and the size is 720,480

so any feedback thanks

[Xesdeeni]

Your PC screen is progressive. TV is interlaced. The conversions we are talking about are designed for TV viewing. For PC viewing, you need to deinterlace, or you will see the interlacing lines. If you are just previewing the video on the PC before burning for the TV, then there is nothing wrong--leave the video interlaced. If your end goal is only the PC, you should deinterlace, and you can even use a different framerate. If your end goal is both the TV and the PC, then you should leave the video interlaced, but use a player (like PowerDVD) on the PC that deinterlaces during playback. If your end goal is a progressive TV, you should still leave the video interlaced and allow the TV to deinterlace when you play it back for maximum quality.

I have a movie trailer and its encoded in

4 : 3
20 fps
PAL
Interlaced

There is no such thing as a 20 fps DVD. But I have found if you check "Force Film" in DVD2AVI for a PAL DVD, you get 20 fps. "Force Film" should only be used for telecined NTSC video, so check "None" instead.

Xesdeeni

[Bluesilo]

Quote:
I have a movie trailer and its encoded in

4 : 3
20 fps
PAL
Interlaced

Bluesilo Posted: Jun 03, 2003 17:05

--------------------------------------------------------------------------------

Nvm bout the 20 fps

Once I output to d2v and load into Dvd2Avi again it reads 25 fps

Anything special I have to do for interlace PAL to interlace NTSC

I believe the file is svcd not sure....its a trailer so


__________________________________________________ ___________
@Xesdeeni

Anyways, Thanks I never realized that pc was progressive and tv was interlace and I guess it didn't help I was playing the newly encode film in Real One Player.....Ill remember to load into PowerDVD thanks

[repdetect2]

Lots of information here, I think I got lost...

I have a PAL movie that DVD2AVI shows as being progressive with AC3 audio.

Can you tell me steps I should take to convert it to NTSC DVD?

I would love a guide on this

Thanks so much.

[DJRumpy]

In DVD2AVI, ensure Force Film is DISABLED. You do not need this for progressive pal sources. Make sure the Audio option is set to 'Decode'. This will give you the WAV which you can convert in BeSweet. Select FILE | SAVE PROJECT. Give the .D2V file any name (we'll call it VIDEO.D2V for this example ). You should also have MPEG2DEC (or MPEG2DEC3 for AVISynth 2.5). This will allow your AVISynth scripts to read DVD2AVI project files.

It's easiest with AVISynth. Create a new text document, and open it in Notepad. Your script would look like the script below. Of course you'll have to correct the path to your mpeg2dec.dll if it's not in your path, or plugin directory. If it's not, you can specify the full path (i.e. "c:\temp\blah\mpeg2dec.dll".

LoadPlugin("mpeg2dec.dll") #make sure the DLL is in your plugin directory
mpeg2source("VIDEO.D2V")
AssumeFPS(23.976,True)

Save this script as VIDEO.AVS (or any name you like for that matter). Put it in the same directory as your VIDEO.D2V file. You can now open the .AVS file in whatever encoder you use. It's output will be 23.976 frames per second.

Convert the WAV file that DVD2AVI generated with BeSweet (select the PAL to FILM (23.976 fps) option. The output WAV can be fed into your encoder as a WAV, or you can plug it right back into your .AVS script using the WAVSource command (of course this assumes you convert your WAV to NTSC, BEFORE dropping your .AVS file onto your encoder. The script with included WAV would look like this:

LoadPlugin("mpeg2dec.dll") #make sure the DLL is in your plugin directory
video=mpeg2source("VIDEO.D2V")
audio=wavsource("yourdvd2aviconvertedwav.wav")
AudioDub(video,audio)
AssumeFPS(23.976,True)

This script will output a completely converted 23.976 video, which can be encoded with no PAL/NTSC conversion options needed, since it's all been handled by AVISynth, and the wav by BeSweet.

If you want to verify that it looks right, just drag/drop your .AVS script onto Media Player (or associate it with media player if you like). The AVS script will play as if it was an AVI, converting your video to NTSC in realtime.

[Xesdeeni]

OK DJRumpy. I've run a test and found that BeSweet definitely does alter the pitch. See for yourself: http://www.geocities.com/xesdeeni2001/index.html. For fairness, try the BeSweet version first, and even before hearing the original, you should notice it seems high. Then note that the AVISynth version, done using the older methods from my page (http://www.geocities.com/xesdeeni2001/StandardsConversion/StandardsConversion-1.html) sound exactly the same, but the process is a bit simpler without BeSweet. Then listen to the WSOLA version.

Granted, there are a lot of problems with WSOLA as a program that need to be addressed (crashes and memory problems), but the algorithm seems to be quite good.

BTW, why do you need the AVISynth script to switch between 23.976p and 25p? TMPGEnc handles that itself (see above page).

Xesdeeni

[repdetect2]

DJRumpy:

Thanks alot Sir!!! I am going to try this afternoon, I will let you know how it goes.

Thanks again.

[DJRumpy]

Unfortunately, the BeSweet authors site is down, and his/her mailbox is full. I'll take your word for it. I trust you .

I'm just quoting what I read, and if what your saying is true, then it offers no benefit over the other speed up/slow down methods. I use an external auditor called Samplitude, which does timestretch without changing pitch. I guess I should get back to basics more often ( I finally upgraded my AVISynth docs after over a year of updates... )

I use the AVISynth command, because I don't use TMPGenc to encode (CCE User). Although you could do this in TMPGenc, this method will work for any encoder that accepts frameserved input, requiring no addtional (specific) instructions outside of the BeSweet conversion.

repdetect2, remember to convert your audio from pal to film first, before messing with the AVS script. That way you can feed the AVS script the WAVSource and AudioDub commands with the converted WAV file (the last script example).

[FulciLives]

A note to DJRumpy and repdetect2

If you are going to use an AVS script as DJRumpy says then you should probably add this line to it:

LanczosResize(720,480)

This will resize the PAL source to the proper D1 resolution of NTSC
If the source is a SVCD then you might want to use half D1 in which case resize to 352x480 instead of 720x480

- John "FulciLives" Coleman

P.S.
If you have an AC-3 file as an audio source please note that BeSweet can transcode from AC-3 to AC-3 while doing the PAL(25fps) to NTSC(23.976fps) frame rate change. If you will be making a final DVD-R or DVD+R etc. then that would be the way to go VS using a WAV to MP2 process.

[DJRumpy]

In general, you should use Bilinear to reduce in size, and bicubic to enlarge. The Lanczos will sharpen you output quite a bit (even more so that Bicubic), which can be undesirable for size reduction. Low bitrate formats like VCD, or SVCD can have a hard time encoding all of those sharp edges. Try a sample to be sure your happy with the output. If it's too sharp or you output has macroblocking on high motion scenes, then try BilinearResize:

BilinearResize(720,480)
or
BilinearResize(480,480)

The LanczosResize option may only be available on AVISynth versions at or above 2.5x. I've never verified it works on 2.0

FulciLives, do you know for sure?

[FulciLives]

In general, you should use Bilinear to reduce in size, and bicubic to enlarge. The Lanczos will sharpen you output quite a bit (even more so that Bicubic), which can be undesirable for size reduction. Low bitrate formats like VCD, or SVCD can have a hard time encoding all of those sharp edges. Try a sample to be sure your happy with the output. If it's too sharp or you output has macroblocking on high motion scenes, then try BilinearResize:

BilinearResize(720,480)
or
BilinearResize(480,480)

The LanczosResize option may only be available on AVISynth versions at or above 2.5x. I've never verified it works on 2.0

FulciLives, do you know for sure?

I have AVISynth 2.5x version installed I think from the newest GORDIAN KNOT package but of course it didn't like Xeesdeeni's SmoothDeinterlacer so I downloaded version 2.08 from the download section of the DOOM9 website and after installing it I was able to use SmoothDeinterlacer as well as LanczosResize but I don't know if it works because I had first installed 2.5x or if all you need is a 2.08 install.

- John "FulciLives" Coleman

P.S.
I'm doing mostly 720x576 to 720x480 so am I "OK" with using LanczosResize or should I try Bilinear and/or Bicubic?

*** EDIT ***
I don't know if this makes a difference but I use TMPGEnc with a DC COMPONENT PRECISION of 10-bit (not 8-bit or 9-bit) and for MOTION SEARCH PRECISION I normally use the second highest setting ... the one that says, "High quality (slow)"

[DJRumpy]

For any high-bitrate format like DVD, you shouldn't need to worry. It will handle the sharp edges without a problem. I would only look at bilinear when going from DVD to SVCD or lower resolutions, and even then, it's always a good idea to test a high motion scene to see if you can get away with the LanczosResize. I also like sharp images (much like yourself I'm guessing ).

Regarding the Motion Precision setting, this setting also takes up additional bitrate, the higher you set the precision. Most store bought DVD's use a 9 (although 10 should be fine for DVD if you want to use it considering the bitrange it has available). For low bitrate formats like VCD, and SVCD, I would leave it at 8. The motion search setting your using is pretty much what most everyone uses. Everyone complains that the slowest setting doesn't produce a visible difference. I'd stick with the one your using.

[FulciLives]

Regarding the Motion Precision setting, this setting also takes up additional bitrate, the higher you set the precision. Most store bought DVD's use a 9 (although 10 should be fine for DVD if you want to use it considering the bitrange it has available). For low bitrate formats like VCD, and SVCD, I would leave it at 8. The motion search setting your using is pretty much what most everyone uses. Everyone complains that the slowest setting doesn't produce a visible difference. I'd stick with the one your using.

So 9-bit is good enough as 10-bit seems to be overkill ... correct?

I had used the HIGHEST (VERY SLOW) setting for the MOTION SEARCH PRECISION in the past but damn it takes FOREVER on my slow ass computer and after reading many people who claimed there was no real visual difference I too starting using the second highest setting ... with a noticeable drop in encoding time (though still slow on my computer boy do I need to upgrade hehehe).

I suppose my encoding time will decrease if I use 9-bit vs 10-bit for the DC COMPONENT PRECISION?

- John "FulciLives" Coleman

[wen]

ok, on the topic of converting pal dvd to ntsc dvd... i have a pal dvd with nice menus (animated) and i wish to convert everything to ntsc. is it possible to retain the menus somehow? like convert the menus separate from the actual movie and then somehow put it back in vob format without having to reauthor (ie. create my own menu and chapters)? if this is somehow possible, please help me out. thanks.

[FulciLives]

ok, on the topic of converting pal dvd to ntsc dvd... i have a pal dvd with nice menus (animated) and i wish to convert everything to ntsc. is it possible to retain the menus somehow? like convert the menus separate from the actual movie and then somehow put it back in vob format without having to reauthor (ie. create my own menu and chapters)? if this is somehow possible, please help me out. thanks.

I have no such experience with extracting menus although I suppose it is possible. Nice menu designs are nice but really they don't make the feature or the extra's any better. Sort of like the packaging ... don't need that either to enjoy the movie.

- John "FulciLives" Coleman

[DJRumpy]

Some of the one-click DVD copy programs promise that sort of functionality, although I've never tried them. Maybe someone who has can post.

If you wanted to take the hands on approach: Assuming this is a dual-layer DVD, which is over 4.3GB, then you would need to manually extract each animated menu, re-encode the main movie using a lower bitrate, and then put all of the menu's back together using something like Sonic Scenarist. Not worth the effort. Easier to just buy the DVD.

[Xesdeeni]

I'm just quoting what I read, and if what your saying is true, then it offers no benefit over the other speed up/slow down methods.

Check out the samples! Anyway, I opted not to use BeSweet in my conversion and use AVISynth instead. Although I didn't need to use AVISynth for the frame conversion (as mentioned above), I used it to convert the audio. I could have used BeSweet and left AVISynth out altogether, but then I'd have to create a second WAV file as an input to TMPGEnc, while AVISynth would just frameserve the audio (did I say "frameserve the audio?").

I use an external auditor called Samplitude, which does timestretch without changing pitch.

HEY! GIVE IT UP! Is that a PD program or commercial? I spent a lot of time looking for full-fledged programs that could do this before I had to settle for that abortion of a program, WSOLA. If you've got one, SPILL THE BEANS!

I use the AVISynth command, because I don't use TMPGenc to encode (CCE User).

OK, so for all you lurkers, if you use TMPGEnc, you don't HAVE to use AVISynth. For CCE or most other encoders, you will.

The LanczosResize option may only be available on AVISynth versions at or above 2.5x. I've never verified it works on 2.0

FulciLives, do you know for sure?

Yes, LanczosResize() is in starting at least as far back as 2.0.7. For DVDs, I prefer this. For VCDs, I prefer bilinear (I actually don't do VCDs much now that I burn DVDs, but I do VCD-res DVDs for extended content, which amounts to the same thing).

@wen

ok, on the topic of converting pal dvd to ntsc dvd... i have a pal dvd with nice menus (animated) and i wish to convert everything to ntsc....

I tried posting the same question on Doom9 a while back (http://forum.doom9.org/showthread.php?s=&threadid=50052) without much luck. The only way I know is to hand-convert all the videos and stills and then hand-author the conversion from scratch. I have one PAL DVD with some complicated animated menus that I keep putting off, because I don't want to do all that "manual labor."

Some of the one-click DVD copy programs promise that sort of functionality, although I've never tried them. Maybe someone who has can post.

I haven't seen one that does standards conversion, only reduces the menus' bit rates for the same standard. Standards converting the menus requires adjustment of not only the images, but the button hot spots, and perhaps other stuff I asked above in the above (unanswered) thread.

Xesdeeni

P.S. How do you guys get the "Xesdeeni wrote:" for quotes?

[DJRumpy]

The Samplitude software is commercial release, and extremely expensive. PM me for details if you want them, as I can't discuss them in forum. It's primary function is to mix various audio tracks. It does timestretching and pitchstretching (essentially changing BPM), all sorts of audio mixer effects, too numerous to list here. I use it to mix Dance/Club music onto CD-R.

I'm curious about the 'Quote' bit too. I never noticed that some actually refer to the original poster, while others just say "Quote:".

[Xesdeeni]

Yeah, I have some expensive software that will do it too (SoundForge), but for my web site, I wanted something free or very cheap. That's why I had to settle for WSOLA, and even that I had to work with to get even a little working. I wish I had the time to work on it more. But since I don't, I wish someone else would grab the source and take a stab.

Xesdeeni

[FulciLives]

Here is how you do the quote thing ...

blah blah blah

I check marked "Disable BBCode in this post" so you could see exactly what the code looks like.

As for WSOLA ... like I said I tried it and got some error about memory or something ... it wouldn't process the audio for me. So for me I use BeSweet and I can't hear no pitch difference ... not saying it is or isn't there ... just that I don't hear it so I'm happy with it.

Check out my post on the "widescreen to fullscreen" thread hehehehe

- John "FulciLives" Coleman

P.S.
I have no clue how to rip menu screens (motion or not) from a DVD ... any guides on this? I'm guessing no

[Xesdeeni]

Here is how you do the quote thing ...

[qute="Xesdeeni"]blah blah blah[/qute]

I could have sworn I tried that. Maybe I forgot the quotes.

As for WSOLA ... like I said I tried it and got some error about memory or something ... it wouldn't process the audio for me. So for me I use BeSweet and I can't hear no pitch difference ... not saying it is or isn't there ... just that I don't hear it so I'm happy with it.

Did you try the samples I uploaded above? Again, for fairness, listen to the BeSweet and AVISynth ones first and see if they sounds high before you compare it to the original or WSOLA ones.

I have no clue how to rip menu screens (motion or not) from a DVD ... any guides on this? I'm guessing no

I haven't seen a guide per se, but the format is pretty much the same as the videos. Still menus are just grouped as separate frames. I just extract the frames with DVD2AVI and save the bitmaps to scale and use as the background. But it's a severe pain.

Xesdeeni

[FulciLives]

Did you try the samples I uploaded above? Again, for fairness, listen to the BeSweet and AVISynth ones first and see if they sounds high before you compare it to the original or WSOLA ones.

No I didn't and perhaps I will do that ... but my point is ... WSOLA didn't work for me ... so ... it really isn't an option then is it?

- John "FulciLives" Coleman

[Xesdeeni]

I assume not many of the tools we use, especially AVISynth, and especially when we tried to use them together, worked the first time. But if the results are worth it, we may put up with the ideosyncracies and try a few times before we give up. With more info, I can probably help you get WSOLA working, just like you've probably helped a few people use other tools. That is of course only if the quality is worth the trouble. That's why I'm curious what you think of the samples.

Xesdeeni

[illini19]

I have done many SVCD PAL to NTSC DVD and used BeSweet to convert the audio, have converted WAV--->AC3 and MP2. I know it has to adjust the pitch to convert from PAL 25 fps to NTSC 23.976 fps but I never heard any difference. I saw your test results and it's interesting.

Just wondering if there is any difference to using AVISynth (loading in Vdub or WSOLA?) then BeSweet even with the stats you showed.

Like I said, I can never hear any difference but we all know it has to adjust the pitch to change the framerate.

Interesting to say the least.

[wen]

just a question... someone online told me to use cce instead of tmpgenc. anyone do this? supposedly cce does a much better job.

[DJRumpy]

illini19, do you mean BeSweet, or just any audio editor in general changing length (framerate) of audio? Length can be changed, while retaining pitch. What they do, is compress the audio, to close out the empty spots in the audio. If you take a wav of someone talking for intance, and magnify it to a large degree, you'll see the wave it produces. In this wave, are bits of data, with gaps between them. What they do, is close the distance in these gaps, to reduce the length, whithout modifying the bits of data containing actaul sound. I trust Xesdeeni regarding BeSweet, but there is software which will timestretch while retaining pitch. I use it myself.

The reverse is also true, where the gaps are enlarged to increase the playtime. Occasionally you will have audio that is so 'full', that the gaps are almost non-existant. In these cases, the compress will actaully chop bits of audio out in extremely small increments, which can be visually heard if the chops get to be large enough, but even that will not change the pitch of the sound.

wen, you should post your own thread if you want that answered by anyone other than those participating in this thread. It's considered rude to hijack a thread lilke your trying to do. In my opinion, yes, CCE is better and faster, bet then again, opinions are like arseholes. Everyone's got one

Use the 'New Topic' button in any forum, or just do a search. There are hundreds of posts regarding the strengths and weaknesses of TMPGenc, CCE, MainConcept, Panasonic, etc. All of these encoders are used by the members here. Do a search to see the arguments.

[Xesdeeni]

@DJRumpy

I'm pretty sure the process you describe is not the process used. The operation you describe is a "time domain" operation. It just removes audible gaps or pauses, which can shorten the audio. However, this will not work with continuous sound, like music. Moreover, even with speech, the results won't sync with any version of the video (either before or after conversion). And just removing samples in the middle of an active waveform will change the pitch, while introducing audible glitches and other artifacts.

There are several techniques used to actually change the speed of audio without adjusting the pitch. A related alteration is to change the pitch without changing the speed. I can't claim to be an expert on the specifics, but all the techniques I've researched work in the "frequency domain."

An example of how this might be done:

Using one of several methods (Fourier Transform, Cosine Transform, etc.) the time domain version of the audio is converted to a frequency domain version (this is the same first step for most audio compressors as well). In this format, each entry (called a bin) contains a magnitude of a particular frequency. You can picture the graphical display on an audio equalizer, where each bar represents the amplitude of a particular frequency. To adjust the pitch without changing the duration, you simply shift the frequencies from one bin to another. If you converted back to the frequency domain at this point, you would have pitch shifted audio with no speed change. To accomplish the speed change, you would then just resample. This would change the pitch, reversing the above pitch change, but the speed would have changed.

In actuality, the algorithms are more involved than I explain above. The shifting of the bin contents is more complex (for a variety of reasons), and even the specifics of how the conversion between domains is done varies widely. Also, I'm pretty sure WSOLA does all the above in the frequency domain, and I also don't think it's actually a two-step process. But the example above should help understand what's going on.

All of the algorithms have their own share of problems. Most of the ones I found sound "swishy" after the conversion. You get something like an echo after a quick sound, and you also get a pre-echo before the sound. WSOLA seems to be much better than the others. The author of the source I used claimed there are no artifacts, but I think they are just very minimal.

Xesdeeni

[DJRumpy]

It sounds like your describing a move from one frequency to another (say 44Khz to 45Khz, and then slowing down the audio to reach the 44Khz sample frequency again?

[Xesdeeni]

Well, not 44KHz, but say you have a WAV file containing a 2KHz tone. When you convert to the frequency domain, you'll have data that has 0's in all but the 2KHz bin, where it will contain the amplitude of the tone (we're ignoring phase for this discussion). Then if you shift this amplitude from the 2KHz bin to the 2.2KHz bin, and then convert back to the time domain, you'll have a 2.2KHz tone. But the duration of that tone will be the same as the duration of the original 2KHz tone. If you just resampled at 110%, you'd have a 2.2KHz tone, but it would be 10% shorter than the original.

Xesdeeni