XVID to KSVCD audio sync problems

[Mildew]

I've got a 90 minute XVID avi (ripped from a DVD).
VIDEO: mpeg4, 576x320, 23.98 fps. AUDIO: mp3, 48000 Hz, stereo, 116 kb/s

Using ffmpegX's preset, I converted it to a KSVCD. The XVID file converted without any errors. I viewed the resulting mpeg file. At the beginning, the audio is in sync, but about 30 minutes in, the audio starts to go out of sync and get progressively worse.

I've got the latest DivX codec, along with XVIDDelegate.component, and the 3ivX codec.

Any idea what's causing this and how I can fix it?

TIA.

[TugBoat]

Any idea what's causing this and how I can fix it?.

Nope. But just for S&Gs, try the work around that I use for DVDs. Demux the MPG and save the Video. Open the XviD in QT Pro and export as WAV. Open the WAV in ffmpegX and select the 'Audio file to mp2' preset. That's assuming you want a mp2. You might want to open the two mp2s in QT and see if there's any difference in lengths. The QT file should be longer. In the Tools tab, enter the video file and the new mp2. Select 'Mux as XSVCD', I have no idea what the difference between KSVCD and XSVCD is. Hit 'Mux as...' and see what happens. It's worth a try 8)

[major]

If you encoded to NTSC FILM framerate you probably also need to enable 3:2 (pulldown) in the Options tab.

[Mildew]

If you encoded to NTSC FILM framerate you probably also need to enable 3:2 (pulldown) in the Options tab.

Okay. I'll try that. BTW, I tried converting to CVD and it came out perfectly.

How do I know when to enable 3:2 pulldown?

[major]

3:2 pulldown should be enabled when encoding NTSC FILM MPEG-2 for hardware playback.

[Mildew]

3:2 pulldown should be enabled when encoding NTSC FILM MPEG-2 for hardware playback.

Is there a reason why ffmpegX doesn't automatically enable 3:2 pulldown when you select the convert to KSVCD preset if the video file (e.g., divX avi) has a NTSC film framerate?

[major]

The reason is a technical one, eg the preset is not a stored selection, but this will be added in future versions.

[TugBoat]

3:2 pulldown should be enabled when encoding NTSC FILM MPEG-2 for hardware playback.

But what about those of us that strictly use 29.97? Apple's MPEG2 encoder does 29.97. DVDSP likes 29.97, so all my encodes are 29.97. I run into this 1/2-2 second sync problem a lot. Last night....I had a PC generated PAL disc that I wanted to convert to NTSC.

I ended up using 0SEx to rip streams because ffmpegX couldn't recognize the VIDEO_TS folder and Yade X had problems. Selected the m2v and added the AC3. The MPG that ffmepg created began in sync but progressively lost it to where it was off by nearly two seconds at the end of the 2'26" movie. I demuxed, and imported to DVDSP 2.05. Same same. I deleted the demuxed AC3 and imported the same ripped AC3 that I had used to encode. PERFECT sync.

Let me repeat that. I used the exact same AC3 file to encode with, using Passthrough/Add and it was out of sync. The exact same AC3 file that wasn't Passed through synced up perfectly. STRANGE!

[major]

NTSC is not NTSC FILM. 3:2 pulldown should not be used with NTSC. You should always use the same framerate as your source. Also, the sync issues come from video being offsync, almost never from audio being offsync.

[Mildew]

NTSC is not NTSC FILM. 3:2 pulldown should not be used with NTSC.

What's the difference? I thought NTSC and NTSC Film were the same thing.

[major]

NTSC is 29.97 frames per second, while NTSC FILM is 23.976 frames per second.

[TugBoat]

No, I never use the 3:2 when encoding at 29.97.
OK, it was late last night when I finished, so, I went back over a few things. First, the MPG is in sync when played with VLC and MPEG Streamclip. QT, of course, doesn't play the audio as it's AC3.

Here's where it starts getting strange. If I open the MPG in ffmpegX it says,
Video: mpeg-2, size: 352x480, 29.970 fps, 9000 kbps,
Audio: ac3, 48000Hz, 448 kbps
But, when I use the tools section to demux I get 15 pairs of files named things like 'ID.mpg-0.m2v, ID.mpg-0.mp3', etc. MPEG Streamclip will demux it properly, but, A.Pack says the AC3 is 2:25:58:14. The original AC3, the one not muxed during encoding, reads 2:26:00:07. Mmmmmm, that's nearly 2 seconds...the amount that the sync is off in DVDSP.

Something else I've failed to figure out is why Audio clips opened in DVDSP always read lower than they do when viewed in the QT Player. Peak, and A.Pack. But, I really don't care what they read, as long as they sync with the Video.

The 15 pairs of demuxed files happens sometimes but that is not the only time the Audio is shorter than it should be to sync. Most of the time ffmpegX properly demuxes the MPGs but the Audio can still fail to be in sync in the MPG or DVDSP.

Maybe the fact that I have to use iTunes to convert the mp2 to AIF because DVDSP will not accept ffmpeg mp2s has something to do with it. Nope that can't be it, 'cause if I select AC3, some still don't sync. Which is why I usually just use QT to export an AIF and send it to A.Pack...

I think I'm rambling.....
:P

[major]

No, I never use the 3:2 when encoding at 29.97.

The whole point is, whether you should encode to NTSC FILM instead (this depends from the framerate of your source).

[TugBoat]

I found out why Yade had problems. Apparently whatever PC program was used on this DVD likes to leave errors that DVDSP hates as well. It's also become apparent that errors in the original get passed through ffmpeg to the output to go along with the ones left by ffmpeg. Great. The single clip 2':26" m2v wouldn't make it past 9.5 minutes before killing DVDSP with the muxer_internal_error. Back to the drawing board.

So, I muxed the original ripped files and passed the mpg through MPEG Streamclip by converting it to a TS, splitting it, then converting back to mpg. I ended up with two mpgs that played fine in everything I tried it with. I had tried splitting in ffmpegX but it kept splitting at 8 minutes instead of 70 minutes.

Then I decided to try the 23.98 framerate and found that you can only get 23.98 if you select 'decode with QT'. Is this normal? I had done both clips before I realized that I wasn't getting 23.98. I had the same problem with the ac3 file. The encoded version was shorter than the original and wouldn't sync. I ran those two clips through DVDSP and got halfway through the second one before another muxer error killed it.

So another encode of the two mpgs. This time I used 'decode with QT' and actually got 23.98. I also got a much lower bitrate than the 2000 I had set. It's closer to 1500. The color is much more saturated as well. Not sure I like that. And apparently MPEG Streamclip reversed the tracks on the first clip so I ended up without audio on that one. Back to the no reversed track option with 'decode with QT' thing. Then you have to run it through the tools section to get 3:2 pulldown. And of course, no manual letterboxing with decode with QT. mmmmm

The second clip wasn't reversed. I dunno. Unfortunately ffmpeg gave me a '**ERROR: [mplex] Too many frame drops -exiting' error on muxing. It did leave the .copy file which works fine as an AC3 once you change the extension. Guess what, the original ac3 from the MPEG Streamclip mpg says 01:17:34:01, the ac3 from the encode says 01:17:34:00. Wish I knew if this was due to the 23.98 framerate or the 'decode with QT' option 8)

I ran the two 23.98 clips through DVDSP using the ac3s from the two Streamclip mpgs and it finally muxed without error. Lotsa hoops. So, it appears that this particular project was hampered by the garbage in, garbage out factor. However, it does appear that encoding at 23.98 or 'decode with QT' may solve the shortened audio problem. I'll have to do some more testing. The simple QT export as AIF is looking pretty good at this point for those movies that don't have 5:1 audio that you want to save.

Thank god all projects aren't like this one :P

[major]

The audio is never shortened, rather it's the video that gets an incorrect duration.

[TugBoat]

It certainly appears to be shortening the audio. I can open the original audio file and it's reads a certain duration. Open the audio file after the ffmpeg encode and it's shorter. What else can explain this?

Should you have to use 'decode with QT' with ffmpeg to get 23.98?

[major]

Perhaps an audio sampling rate conversion?

And you can get 23.976 with any encoder, the only thing QT deals with better, is normalizing framerate in case of a framerate change.

[TugBoat]

I'm receiving shortened audio from AC3s that are being passed through, using 'Add Audio', at the same sampling. The last two DVDs I've tried have had this problem. The one I'm working on now is worse than the last one, even the mpg that ffmpeg outputs is out of sync. The Audio slowly creeps ahead of the video. This one has me totally baffled, the DVD is in sync, demux then remux the exact same files and it's out of sync. I've never had this problem before. Looks like I'm going to have to try something new, it looks like encoding each VOB separately may work.
I've also tried using the same rate on earlier AVIs without success. It must be something else.

The framerate problem seems to be with MPEG as input. Try using an MPEG2 as input/output, with the ffmpeg engine. When I select 23.98, without decode with QT, the output is 29.97. If I simply select decode with QT, the output is 23.98. Happens everytime for me. I tried it without MPEG as input and it works fine. Until I tried some more AVIs. So it seems you can't, for the most part, export at 23.98 with ffmpeg without using 'decode with QT'. I like manual letterboxing & cropping, it looks like it's 29.97 for me

[TugBoat]

Here's my experiences with the last 4 AVIs that I converted to MPEG2 to make a DVD in DVDSP 2.05. I used the ffmpeg codec set at half DVD, 29.97, manual letterboxing and cropping.

Input #0, avi, from '/Volumes/Video/Movies/Movie1.avi':
Duration: 00:58:29.6, bitrate: 1672 kb/s
Stream #0.0: Video: mpeg4, 640x288, 25.00 fps
Stream #0.1: Audio: mp3, 48000 Hz, stereo, 128 kb/s
Output #0, vob, to '/Volumes/Video/Movies/Movie1.avi.ff.mpg':
Stream #0.0: Video: mpeg2video (hq), 352x480, 29.97 fps, q=1-8, pass 1, 1600 kb/s
Stream #0.1: Audio: mp2, 48000 Hz, stereo, 128 kb/s
Stream mapping:
Stream #0.0 -> #0.0
Stream #0.1 -> #0.1
[mpeg2video @ 0x331210]rc buffer underflow
[mpeg4 @ 0x331210]vop not coded
bench: utime=2129.920s
video:564649kB audio:54837kB global headers:0kB muxing overhead 21.150753%
Input #0, avi, from '/Volumes/Video/Movies/Movie1.avi':
Duration: 00:58:29.6, bitrate: 1672 kb/s
Stream #0.0: Video: mpeg4, 640x288, 25.00 fps
Stream #0.1: Audio: mp3, 48000 Hz, stereo, 128 kb/s
Output #0, vob, to '/Volumes/Video/Movies/Movie1.avi.ff.mpg':
Stream #0.0: Video: mpeg2video (hq), 352x480, 29.97 fps, q=1-8, pass 2, 1600 kb/s
Stream #0.1: Audio: mp2, 48000 Hz, stereo, 128 kb/s

The demuxed AIF from the ffmpeg MPG has a duration of 00:58:28:27
The QT AIF exported from the AVI has a duration of 00:58:29:18
The QT AIF syncs in DVDSP fine, the ffmpeg one is noticeably off.


Input #0, avi, from '/Volumes/Video/Movies/Movie2.avi':
Duration: 00:57:24.2, bitrate: 1664 kb/s
Stream #0.0: Video: mpeg4, 640x288, 25.00 fps
Stream #0.1: Audio: mp3, 48000 Hz, stereo, 128 kb/s
Output #0, vob, to '/Volumes/Video/Movies/Movie2.avi.ff.mpg':
Stream #0.0: Video: mpeg2video (hq), 352x480, 29.97 fps, q=1-8, pass 1, 1600 kb/s
Stream #0.1: Audio: mp2, 48000 Hz, stereo, 128 kb/s
Stream mapping:
Stream #0.0 -> #0.0
Stream #0.1 -> #0.1
video:547483kB audio:53816kB global headers:0kB muxing overhead 21.875515%
bench: utime=2039.100s
Input #0, avi, from '/Volumes/Video/Movies/Movie2.avi':
Duration: 00:57:24.2, bitrate: 1664 kb/s
Stream #0.0: Video: mpeg4, 640x288, 25.00 fps
Stream #0.1: Audio: mp3, 48000 Hz, stereo, 128 kb/s
Output #0, vob, to '/Volumes/Video/Movies/Movie2.avi.ff.mpg':
Stream #0.0: Video: mpeg2video (hq), 352x480, 29.97 fps, q=1-8, pass 2, 1600 kb/s
Stream #0.1: Audio: mp2, 48000 Hz, stereo, 128 kb/s

The demuxed AIF from the ffmpeg MPG has a duration of 00:57:15:10
The QT AIF exported from the AVI has a duration of 00:57:24:07
The QT AIF syncs in DVDSP fine, the ffmpeg one is noticeably off.

Input #0, avi, from '/Volumes/Video/Movies/Movie3.avi':
Duration: 02:03:47.0, bitrate: 790 kb/s
Stream #0.0: Video: mpeg4, 608x320, 25.00 fps
Stream #0.1: Audio: mp3, 48000 Hz, stereo, 128 kb/s
Output #0, vob, to '/Volumes/Video/Movies/Movie3.avi.ff.mpg':
Stream #0.0: Video: mpeg2video (hq), 352x480, 29.97 fps, q=1-8, pass 1, 1400 kb/s
Stream #0.1: Audio: mp2, 48000 Hz, stereo, 128 kb/s
Stream mapping:
Stream #0.0 -> #0.0
Stream #0.1 -> #0.1
[mpeg2video @ 0x331210]rc buffer underflow
bench: utime=4431.260s
video:1043134kB audio:116047kB global headers:0kB muxing overhead 21.020850%
Input #0, avi, from '/Volumes/Video/Movie3/Movie3.avi':
Duration: 02:03:47.0, bitrate: 790 kb/s
Stream #0.0: Video: mpeg4, 608x320, 25.00 fps
Stream #0.1: Audio: mp3, 48000 Hz, stereo, 128 kb/s
Output #0, vob, to '/Volumes/Video/Movies/Movie3.avi.ff.mpg':
Stream #0.0: Video: mpeg2video (hq), 352x480, 29.97 fps, q=1-8, pass 2, 1400 kb/s
Stream #0.1: Audio: mp2, 48000 Hz, stereo, 128 kb/s

The demuxed AIF from the ffmpeg MPG has a duration of 02:03:45:21
The QT AIF exported from the AVI has a duration of 02:03:47:00
The QT AIF syncs in DVDSP fine, the ffmpeg one is noticeably off.

I'm going to stop, well, Movie4 went
The demuxed AIF from the ffmpeg MPG has a duration of 01:52:56:15
The QT AIF exported from the AVI has a duration of 01:52:57:15
Also notice the Movie duration in the second line of each report and how that matches the QT AIFF duration.

It certainly appears that the encoding is shortening the audio and causing sync problems in my case. Fortunately, I don't have a problem using QT Pro to obtain my audio.

Hope this helps.....