Fixing an improperly mastered DVD

[Cronos]

Hello sages,

I've got a set of dvds which apparently have been improperly mastered. While the audio and video are in sync, the playback is about 4% slower than it should be. The result is that the audio pitch is lower for all voices - really irritating. I'm guessing that the original material was at something like 23.976 frames/sec but was mistaken as 25 frames/sec when it was converted to dvd-ntsc at 29.976.

So, any ideas on remastering/converting to correct this problem with a minimal (no) loss of quality? I've tried playing around with some tools but the video encoding always seems to be the issue. And, there is the question of how to get this back into a dvd.

Cronos

[guns1inger]

I would start by confirming exactly what you have. What resolution, what framerate, doe sit use pulldown etc. You will have to know these things before you can plan an attack on it. It may be impossible, or at least very difficult, to IVTC the material if indeed it has been encoded in the manner you describe, or it may be a simple matter some audio tweaking.

Work out what you have, and we can tell you how to change it.

[manono]

And give us a sample to study.

Open a VOB in DGIndex, use the [ and ] buttons to isolate a small 10 second piece that displays motion, go File->Save Project and Demux Video, upload the resulting M2V to a third party hosting service such as MediaFire, and post the link here.

I can understand how the audio could become lower than it's supposed to be, but just haven't ever seen it done on a retail DVD before. I did it myself by mistake a couple of months ago.

[Cronos]

Thanks for the responses.

Here is a short section using DGIndex:

http://www.mediafire.com/?sharekey=0cf70ef028f10c82d2db6fb9a8902bda

Its reporting 720x480 frame size, 29.970030 fps and Film 98%(?), Bottom field order.

Let me know what you think.

Cronos

[manono]

How do you know it's pitched too low? Do you have other films of this series (whatever it is) so you know the voices and can compare? Are you a musician or have perfect pitch and can tell that the musical parts are a semi-tone too low?

Anyway, to fix it without reencoding and lowering the quality, demux the entire thing (PGCDemux, making sure to check the "Demux Video Stream" box and to "Check A/V Delay" for any possible audio delay - also, tick "By PGC" rather than "By VOB ID" ) and run the M2V through DGPulldown set for 25->29.97fps. If the audio is lower pitched then the movement is also too slow, and that will speed up the movement to what it was originally. It will also shorten the video and necessitate reencoding the audio (which was unavoidable anyway). I would first take the demuxed AC3 audio and run it through BeSweet using the Preset for NTSC->PAL (23.976 to 25) to create a 25fps WAV. Then reencode to AC3 using whatever you have, Aften and one of its GUIs if you need freeware, or a good commercial product such as SoftEncode, if you have one. But there are other ways to handle the audio.

Then remux (Muxman, not forgetting to add in the chapters by loading the Celltimes.txt in File->Import Chapters), and stick it back into the DVD using VoBlanker. Although this guide is about adding subs to a DVD, the principle is the same. You'll be following the same steps as in Method 2 (ignoring the stuff about subtitles):

https://forum.videohelp.com/topic338721.html

[Cronos]

Hmmmm...

I tried the DGPulldown from 25->29.97 and that seemed to speed up the video too much. I've compared a clip with one on youtube (search for reboot) and the sped-up m2v soon got out of sync with the baseline there.

On some further research, I've found the following statement:

"However, it was improperly mastered as the 25fps source material was treated as 24fps film speed material, meaning 3:2 pulldown flags were encoded into the mpeg stream which results in the video playing back 4.096% slower and all the voices sounded deeper. Anchor Bay have since corrected and remastered the fourth season disc, but it is only available by contacting them for a replacement."

I think I'll take a stab at getting the replacement disk and put this little project on hold for now. I've tried using a custom input frame-rate to get the 4% but I'm concerned that this will make synching the modified audio problematic. I think that speeding up the video is doable with your advice but the audio is a bit more tricky.

Thanks for your help!

Cronos

[manono]

I tried the DGPulldown from 25->29.97 and that seemed to speed up the video too much. I've compared a clip with one on youtube (search for reboot) and the sped-up m2v soon got out of sync with the baseline there.

Based on that quotation you gave, running it through DGPulldown set for 25->29.97 is exactly the right thing to do, and of course it'll get out-of-synch when compared to one slowed from its true speed. By running it through DGPulldown you speeded it up to its original real speed, an amount equal to about 4% faster when compared to the too-slow DVD speed.

I think that speeding up the video is doable with your advice but the audio is a bit more tricky.

Not tricky at all. I already explained how to do it in some detail. I do similar things several times a week, although I'm usually slowing down speeded up PAL stuff to the original NTSC speed.

But I also agree that going for a replacement disc is the best thing to do. Good luck.

[Cronos]

I must be misunderstanding something here

In the mastering to dvd, the producers thought that the source material was 24 fps. During the conversion to 29.97, the progressive frames were split into fields with some new frames created with fields from different source frames (basically, how I understand the pull down thing works). This added a number of new frames, say X. Since the source was actually 25 progressive fps, they should have added (as above) a number of frames X' < X.

Because more frames were added then needed, the time to play a sequence was stretched to a longer time, hence it plays slower than it should. I'm probably simplifying it too much and not describing this accurately

So, when we do the dgpulldown of 25->29.97, the program reconstructs the original progressive frames (or does this happen in the pgcdemux stage?) and then does the correct interpolation and introduces the new number of frames (X') to get the correct playback.

Is this essentially correct?

For the audio, "run it through BeSweet using the Preset for NTSC->PAL (23.976 to 25)". Wouldn't that be NTSC->PAL (29.97 to 25) or is the audio stream at 23.976 after the demux? So, when I encode to AC3 again, this would take the 25 fps WAV to a 29.97 fps AC3?

Thanks for taking the time to answer my questions. Hopefully I'm not exhausting anyone's patience

Cheers,

Cronos

[guns1inger]

If the source was 24 fps then they should have encoded it at 23.976 fps and applied 3:2/2:3 pulldown so it played back correctly at 29.970 fps interlaced, or 23.976 fps progressive, depending on the output equipment. No new frames need to be created and the audio running time does not get altered. The extra frames are created on-the-fly by the playback unit as required.

Poorly mastered material or material that is also being converted for VHS release on the cheap is mastered by creating the extra frames during the encoding, usually by blending existing fields in unfortunate ways. This results in artifacts and ghosts and is not particularly pretty.

Encoding material with the wrong parameters - e.g. telling the encoder that the source is 23.976 fps when it is really 25 fps - can produce more unfortunate outcomes.

If you can get it back to the original 25 fps source fields, then you have the progressive (or interlaced, depending on the source - PAL can be either) frames. Using DGPulldown's 25->29.970 setting simply inserts hidden flags to tell playback NTSC hardware to create extra frames on-the-fly to fill to get to 29.970 interlaced frames per second for viewing.

[manono]

Because more frames were added then needed, the time to play a sequence was stretched to a longer time, hence it plays slower than it should. I'm probably simplifying it too much and not describing this accurately

That's sort of correct. But no frames were actually added for the NTSC DVD. As you said, they treated the source as 23.976fps and added flagging (software added into the video stream, sort of) that tells the player how to add additional fields to pad the output to interlaced 29.97fps, as required for NTSC DVDs. guns1inger explained it well in his last paragraph. Thus the output framecount was increased by 25% (29.97/23.976=1.25), which, as you said, stretched the video "to a longer time, hence it plays slower than it should". By properly reflagging it as 25->29.97fps, you are indeed stretching the output less (29.97/25=1.1988), in this case by less than 20% which, in effect, keeps it the original length. The question then arises how this remastered DVD is going to look. No retail, commercial DVD has ever, to my knowledge, applied soft pulldown as you did using DGPulldown to go from 25fps to 29.97fps. Sometimes they'll apply hard pulldown (encode the additional fields into the video stream) to do it, which can be OK, but even that's fairly rare. More often they'll field-blend the heck out of it and make a blended, blurry, mess. That is, by doing it yourself there's a very real chance you'll wind up with a better quality DVD than the replacement. I'd be curious to have a sample to examine, if and when you get your replacement.

Wouldn't that be NTSC->PAL (29.97 to 25) or is the audio stream at 23.976 after the demux? So, when I encode to AC3 again, this would take the 25 fps WAV to a 29.97 fps AC3?

Audio doesn't really have a framerate, but a length. The BeSweet setting is just given in framerate terms because it may (or may not) be easier to understand that way. But no, it wouldn't be 29.97->25. Go back to your quotation about what happened with the video. It was treated as 24fps and not 29.97fps. Film is 24fps and not 29.97fps or 30fps. Remember that they added 3:2 pulldown to go from 23.976fps to 29.97fps. By using the 23.976->25fps setting you'll be speeding it up, shortening its length by about 4% and raising the pitch by a semitone to bring it back to normal. The upshot is that if you follow my instructions, you'll wind up with both the audio and video the proper length and in synch.

[Cronos]

Well, I gave it a shot and unfortunately I do seem to have some audio sync problems.

To recap my entire process.

Using DVDDecrypter, I extracted the .vob files to my hard drive.
Using PGCDemux, I extracted to two.ac3 files (2 of them, one is 5.1 Surround and the other was Dolby Surround 2.0) and a single m2v file.
Using DGPulldown, I created a m2v file using the 25 to 29.97 setting.
Using BeSweet, I converted one of the ac3 files to a new ac3 using the '23.976 to 25.000' setting. Listening to the audio, it's definitely better.
Using Muxman (free version), I specified the recoded ac3 and recoded m2v files, loaded the celltimes.txt, and generated some .vob files.

Using PowerDVD, I played the .vob files and everything starts out in sync at the beginning of the movie but jumping to the end, I'm off by several seconds. The video seems to be leading the audio at this point.
PcgDemux reports no audio delay for either of the two audio streams.

I've checked using DGIndex on the demuxed m2v file that clips near the beginning, middle, and end have a general (visual) pattern of 3 full frames and 2 interlaced frames. Setting DGIndex to not honor the pulldown flags produces full frames as expected.

So, what is taking me off the rails here? Could they have done something weird with the audio such as increasing the pitch without changing the length? Is the assumption of a 25 fps original source correct? Hmmm. Well, at least I'm learning

Cronos

[manono]

So, what is taking me off the rails here?

Probably my fault. I was taking them at their word and assuming a real 25fps video, but there may be drop-frame issues involved.

The video seems to be leading the audio at this point.

OK, by the end of the video, the video is a few seconds ahead of the audio? How much it's ahead would depend on the length of the video.

I wish I had this thing on my hard drive as I could get to the bottom of it in no time. Try this; take the original M2V as demuxed by PGCDemux and run it through DGPulldown again, this time checking the "Custom" box and applying pulldown for 24.975->29.97. That will slow the video by a few seconds over its entire length and maybe put it back into synch with the audio. Load everything else into Muxman just as you did before, but this time use the new M2V you created.

[Cronos]

Gave it a shot with 24.975 and it's still not quite right. Oh well...

[manono]

It's closer but still not perfect? It still has a slight progressive asynch - still gets a bit more off as the video goes along? Then without having the entire thing on my hard drive I wouldn't know where the process went wrong, although I suspect the audio now as you seemed to say you did the entire AC3->AC3 23.976->25.000fps in one fell swoop using BeSweet. I'd suggest redoing the audio converting to PCM WAV first at 23.976->24.975 before then using Aften and one of its GUIs to convert the WAV to AC3, but I have no confidence that'll work for you.

I guess back to plan B - get the replacement DVD from Anchor Bay. I'm sorry I wasn't of more help.

[Cronos]

No worries! I definitely appreciate all the suggestions and help you've given so far.

I think I'll just put it on the back burning again. I haven't heard from Anchor bay but I'm still hoping.

Cronos