Thanks for the tip 4partee. I could look for the "AC3" line in mplayer's output, and grab the next number to find out how many channels there are. The -identify option also spits out an ID_AUDIO_NCH value that appears to be number of channels; what happens when you run this command on your 5.1-channel video?

See if the ID_AUDIO_NCH value is 6, or something else.

Also, for deinterlacing, I remembered seeing some quality comparisons a while back. See here. It looks like 'lb' (linear blend) deinterlacing is lower quality in that example; perhaps using 'md' would give you better results, until I can do some more research and add the option to tovid. I believe using linear blend just blends together neighboring scanlines, which is a bad idea since you're really getting an overlay of two different frames. There's also another site that has some postprocessing comparisons; the 'md' deinterlacer looks much better than 'lb' there also. I suspect it's not very noticeable unless you're looking at individual frames, so it depends on whether you want sharpness in each frame, or a kind of "motion-blur."

Actually, I used 'lb' during capture and in tovid. I did not notice any undeinterlacing(that word gave me a chuckle).

Logically, to me, it seems that there would be nil extra processing by leaving in 'lb'. I have not tested but I could run a test. Frankly, I don't give a damn". Sorry, couldn't resist. I just watched GWTW, see?

I understand that mpeg2 is not designed to be edited. I'll stick with lavrec capture since it can be edited by glav. It generates very large output but handles %d in the outfile name to avoid the 2GB avi limit. At 720x480 it needs deinterlacing though as I cant find any lavrec option. Also, I have not tried glav'ing a file that needs deinterlacing either.

As requsted:
mplayer -vo null -ao null -frames 1 -identify /mnt/dvd/sw4anh.avi
<snip>
Opening audio decoder: [liba52] AC3 decoding with liba52
Using SSE optimized IMDCT transform
AC3: 5.1 (3f+2r+lfe) 48000 Hz 448.0 kbit/s
Using MMX optimized resampler
AUDIO: 48000 Hz, 2 ch, 16 bit (0x10), ratio: 56000->192000 (448.0 kbit)
Selected audio codec: [a52] afm:liba52 (AC3-liba52)
==========================================================================
ID_FILENAME=/mnt/dvd/sw4anh.avi
ID_VIDEO_FORMAT=DIVX
ID_VIDEO_BITRATE=819248
ID_VIDEO_WIDTH=720
ID_VIDEO_HEIGHT=480
ID_VIDEO_FPS=29.970
ID_VIDEO_ASPECT=1.7793
ID_AUDIO_CODEC=a52
ID_AUDIO_FORMAT=8192
ID_AUDIO_BITRATE=448000
ID_AUDIO_RATE=48000
ID_AUDIO_NCH=2
ID_LENGTH=7485


The channels are 2 not 6 in the identify output.
=======================================
using:
$ mplayer -vo null -ao alsa -channels 6 -frames 1 -identify /mnt/dvd/sw4anh.avi
<snip>
Opening audio decoder: [liba52] AC3 decoding with liba52
Using SSE optimized IMDCT transform
AC3: 5.1 (3f+2r+lfe) 48000 Hz 448.0 kbit/s
Using MMX optimized resampler
AUDIO: 48000 Hz, 6 ch, 16 bit (0x10), ratio: 56000->576000 (448.0 kbit)
Selected audio codec: [a52] afm:liba52 (AC3-liba52)
==========================================================================

ID_FILENAME=/mnt/dvd/sw4anh.avi
ID_VIDEO_FORMAT=DIVX
ID_VIDEO_BITRATE=819248
ID_VIDEO_WIDTH=720
ID_VIDEO_HEIGHT=480
ID_VIDEO_FPS=29.970
ID_VIDEO_ASPECT=1.7793
ID_AUDIO_CODEC=a52
ID_AUDIO_FORMAT=8192
ID_AUDIO_BITRATE=448000
ID_AUDIO_RATE=48000
ID_AUDIO_NCH=6
ID_LENGTH=7485

Observation:
-identify is reporting what has been requested for output audio processing, not what exists in the audio source. -channels 2 is the default.

Thanks for the 'md' tip. I prefer 'md'. Sadly I can't capture gwtw again!

Also, I have decided to go with two DVD's instead of trying to compress to one.

I wonder if it would work to use -channels 6 for all identification. I just tried it with a normal stereo file, and it still reported AUDIO_NCH=2, presmuably meaning it isn't inserting "fake" channels. I don't know whether ffmpeg can do 5.1 encoding anyway (it can encode multiple channels, but I don't see any options for treating them with 5.1 in mind - the LFE channel should probably go through a low-pass filter of some kind, for example).

However, I don't think this will be especially important, since I plan to just probe for valid AC3 (any number of channels) and copy it as-is if it's there. The 'idvid' script is starting to shape up; I'm using my earlier technique of quickly copying the audio stream to /dev/null to determine duration. I tried several other approaches, but most of them resulted in the same speed. Accuracy is more important than speed, in this case. I'll have to test it on some other videos, to make sure there are no unexpected results.

Also, about the deinterlacing: From what I can determine, all of mplayer's deinterlacing routines cause data to be lost (half the vertical resolution, effectively). I'm assuming it must operate on a frame-by-frame basis, because you would think it would be easy enough to reconstruct the complete frame from the fields present in two successive interlaced frames. Maybe it's not that simple. Anyway, you might also try other deinterlacing routines: run 'mplayer -pphelp' to see a list of them.

edit: Aha... now I see. With interlacing, there are really no "completed frames." Each field is offset from the previous field by 1/60th second, so the only way to get the two fields to "line up", so to speak, is to average or blend them together, thus losing some vertical resolution. This might be worthy of an illustration on the Wikipedia article on Interlacing...

LFE, or Low Frequency Effects, is already low frequency only. Purists say the subwoofer should roll-off beginning at 62cycles. It is not intended for music bass. It is the 'rumble' channel. In setups w/o full range satellites, the subwoofer must be called on to supply music bass and thus is not pure LFE and roll-off needs to be set much higher to fit with the satellites. In my speakers I have full range at the four corners, 12" drivers front and 8" drivers surround, but no subwoofer as it would most likely just intefere. My LFE is redirected to the speakers I designate as 'Large'. At least that is what I have read so far.

Yes, copy.

On deinterlacing , all I know is at 352x240 I have none. Going to read that wiki site.

OK, I wasn't sure about that; so I guess the problem would be that ffmpeg might not be able to determine which of the 6 channels is LFE, were it to re-encode any existing 5.1 AC3. So definitely, everything points to copying existing streams. Until such time as I find a utility to generate 5.1 from regular stereo But that is low on my list of priorities.

A caution to any of you working with VCD: The most recent version of tovid does not produce valid VCD output. It's due to the -q quantization setting, which I added for all formats, including VCD. I didn't think it'd be a problem since the mpeg2enc manual page says "Flag settings that would result in a non-standard stream structure are simply ignored", about its VCD setting. Well, it lies. Having -q in there results in a variable bitrate stream (which is why I was getting bizarre bitrates from idvid and every other program).

So anyway, if you need to encode VCD, stick with version 0.16 for now.

Hrm, lot of techincal discussion going on, but i am reading them all. I pick up things every now and then just from reading posts, even though they are very techinical.

Ah, vcd doesn't work, heh? Just been doing svcd's, so couldn't test that one. And you mean to use tovid 0.15 instead, correct? Or is there a 0.17 release already since you said to use 0.16...

A suggestion to improve 'stats':

The version of tovid is not displayed in .tovid/stats. I think it would be helpful for you for debugging. Probably be helpful for us to since they are easier to sort and compare.

Er, yes. Use 0.15 for VCD. Got 0.17 on the brain, I guess

About putting the version in the ~/.tovid/stats file - the version name should already be printed out if the stats file is not already there. But if you're keeping ongoing stats, then yes, it would be good to put the version in. Maybe just a quick "0.16" at the beginning of the line or whatever. It would be especially good considering the rapid pace of development!

Keep emailing me your stats file occasionally (and anyone else who wants to - wapcaplet99 at yahoo dot com), so I can see what kind of bitrates and file sizes you're ending up with.

stats send.
Yes, sorry, thats what i meant.

Ok, i am trying to use more of the suite's scripts, so the next one i tried is 'makexml' (0.16). This is the command i used:

So, i have a 'xmlout' file, to which i used 'vcdxbuild' like so:

However, i get this error:

vcdxbuild xmlout
xmlout:17: validity error: Element segment-items content does not follow the DTD, expecting (segment-item)+, got
</segment-items>
^
**ERROR: error while parsing file `xmlout': No such file or directory


Am i doing something wrong or is it a problem with the script? All the movies are in svcd format.

Thanks.

Error in the script. The segment-items section is for menus and stills; I neglected to have it check for the presence of any segment-items before inserting the XML for them.

You should be able to just remove the segment-items section from the XML file, and author the VCD normally. Let me know if it works!

Edit: You will also need to remove any lines in the XML that look like this:

With no menu, these will also cause errors from vcdxbuild.

Edit #2: Yeah, you know what, most of the VCD/SVCD output from 'makexml' is broken. I am glad you posted about it sk545, since I had no idea there were so many things wrong! Mostly it's just minor syntactical stuff (since vcdxbuild expects things to be defined in a certain order, which I was not aware of). Anyway, as you said, that's what alpha versions are for I'm getting most of them cleaned up now, so it should be much improved in 0.17. Should be released later this week.

Hmm...i think it would be better for me to wait for the next version of 'makexml' since there are many 'segment' related lines. I will probably mess it up more and waste time.

I did get your 'idvid' email, i will test it out soon and let you know.

Thx.

Edit, here are some preliminary tests with 'idvid', you can compare them to my stats file if you want, since the names are the same in the stats file that i sent you:

./idvid.sh *.mpg *.avi
--------------------------------
idvid video identification script
Version 0.17
Written in 2004 by Eric Pierce
http://tovid.sourceforge.net/
--------------------------------
Gathering video information. This may take several minutes,
so please be patient...
=====================================================
File: jimkarate.mpg
Width: 480 pixels
Height: 480 pixels
Duration: 0:4:19 hours/mins/secs
Framerate: 29.970 frames per second
Video format: MPEG2
Audio format: mp2
Video bitrate: 2500000 bits per second
Audio bitrate: 224000 bits per second
Audio channels: 2 channels
Audio sampling rate: 44100 Hz
=====================================================
Audio is compliant with the following formats:
VCD 1.1
VCD 2.0
2-channel 224000 bps SVCD
Video is compliant with the following formats:
2500000 bps 29.970 fps MPEG2 NTSC SVCD
You can burn this video to NTSC SVCD
=====================================================



--------------------------------
idvid video identification script
Version 0.17
Written in 2004 by Eric Pierce
http://tovid.sourceforge.net/
--------------------------------
Gathering video information. This may take several minutes,
so please be patient...
=====================================================
File: satbashdenoised.mpg
Width: 480 pixels
Height: 480 pixels
Duration: 0:42:56 hours/mins/secs
Framerate: 29.970 frames per second
Video format: MPEG2
Audio format: mp2
Video bitrate: 2500000 bits per second
Audio bitrate: 224000 bits per second
Audio channels: 2 channels
Audio sampling rate: 44100 Hz
=====================================================
Audio is compliant with the following formats:
VCD 1.1
VCD 2.0
2-channel 224000 bps SVCD
Video is compliant with the following formats:
2500000 bps 29.970 fps MPEG2 NTSC SVCD
You can burn this video to NTSC SVCD
=====================================================



--------------------------------
idvid video identification script
Version 0.17
Written in 2004 by Eric Pierce
http://tovid.sourceforge.net/
--------------------------------
Gathering video information. This may take several minutes,
so please be patient...
=====================================================
File: smallinte.mpg
Width: 480 pixels
Height: 480 pixels
Duration: 0:42:26 hours/mins/secs
Framerate: 29.970 frames per second
Video format: MPEG2
Audio format: mp2
Video bitrate: 2500000 bits per second
Audio bitrate: 224000 bits per second
Audio channels: 2 channels
Audio sampling rate: 44100 Hz
=====================================================
Audio is compliant with the following formats:
VCD 1.1
VCD 2.0
2-channel 224000 bps SVCD
Video is compliant with the following formats:
2500000 bps 29.970 fps MPEG2 NTSC SVCD
You can burn this video to NTSC SVCD
=====================================================



--------------------------------
idvid video identification script
Version 0.17
Written in 2004 by Eric Pierce
http://tovid.sourceforge.net/
--------------------------------
Gathering video information. This may take several minutes,
so please be patient...
=====================================================
File: vanilla.mpg
Width: 480 pixels
Height: 480 pixels
Duration: 0:2:24 hours/mins/secs
Framerate: 29.970 frames per second
Video format: MPEG2
Audio format: mp2
Video bitrate: 2500000 bits per second
Audio bitrate: 224000 bits per second
Audio channels: 2 channels
Audio sampling rate: 44100 Hz
=====================================================
Audio is compliant with the following formats:
VCD 1.1
VCD 2.0
2-channel 224000 bps SVCD
Video is compliant with the following formats:
2500000 bps 29.970 fps MPEG2 NTSC SVCD
You can burn this video to NTSC SVCD
=====================================================



--------------------------------
idvid video identification script
Version 0.17
Written in 2004 by Eric Pierce
http://tovid.sourceforge.net/
--------------------------------
Gathering video information. This may take several minutes,
so please be patient...
=====================================================
File: saturday.night.live.presidential.bash.2004.pdtv-lol.avi
Width: 512 pixels
Height: 384 pixels
Duration: 0:42:57 hours/mins/secs
Framerate: 29.970 frames per second
Video format: XVID
Audio format: mad
Video bitrate: 982976 bits per second
Audio bitrate: 128000 bits per second
Audio channels: 2 channels
Audio sampling rate: 48000 Hz
=====================================================
Audio is compliant with the following formats:
Video is compliant with the following formats:
This video does not seem to be compliant with (S)VCD or DVD
standards. If you burn it to a video disc, it may not work.
=====================================================



--------------------------------
idvid video identification script
Version 0.17
Written in 2004 by Eric Pierce
http://tovid.sourceforge.net/
--------------------------------
Gathering video information. This may take several minutes,
so please be patient...
=====================================================
File: smallville.s04e04.hdtv-lol.avi
Width: 624 pixels
Height: 352 pixels
Duration: 0:42:25 hours/mins/secs
Framerate: 23.976 frames per second
Video format: XVID
Audio format: mad
Video bitrate: 1007296 bits per second
Audio bitrate: 128000 bits per second
Audio channels: 2 channels
Audio sampling rate: 48000 Hz
=====================================================
Audio is compliant with the following formats:
Video is compliant with the following formats:
This video does not seem to be compliant with (S)VCD or DVD
standards. If you burn it to a video disc, it may not work.
=====================================================



./idvid.sh -terse *.mpg *.avi

WIDTH:480
HEIGHT:480
DURATION:259
FPS:29.970
VIDEO_FORMAT:MPEG2
AUDIO_FORMAT:mp2
VIDEO_BITRATE:2500000
AUDIO_BITRATE:224000
AUDIO_CHANNELS:2
AUDIO_SAMPLING_RATE:44100
A_VCD1
A_VCD2
A_SVCD
V_SVCD


WIDTH:480
HEIGHT:480
DURATION:2576
FPS:29.970
VIDEO_FORMAT:MPEG2
AUDIO_FORMAT:mp2
VIDEO_BITRATE:2500000
AUDIO_BITRATE:224000
AUDIO_CHANNELS:2
AUDIO_SAMPLING_RATE:44100
A_VCD1
A_VCD2
A_SVCD
V_SVCD


WIDTH:480
HEIGHT:480
DURATION:2546
FPS:29.970
VIDEO_FORMAT:MPEG2
AUDIO_FORMAT:mp2
VIDEO_BITRATE:2500000
AUDIO_BITRATE:224000
AUDIO_CHANNELS:2
AUDIO_SAMPLING_RATE:44100
A_VCD1
A_VCD2
A_SVCD
V_SVCD


WIDTH:480
HEIGHT:480
DURATION:144
FPS:29.970
VIDEO_FORMAT:MPEG2
AUDIO_FORMAT:mp2
VIDEO_BITRATE:2500000
AUDIO_BITRATE:224000
AUDIO_CHANNELS:2
AUDIO_SAMPLING_RATE:44100
A_VCD1
A_VCD2
A_SVCD
V_SVCD


WIDTH:512
HEIGHT:384
DURATION:2577
FPS:29.970
VIDEO_FORMAT:XVID
AUDIO_FORMAT:mad
VIDEO_BITRATE:982976
AUDIO_BITRATE:128000
AUDIO_CHANNELS:2
AUDIO_SAMPLING_RATE:48000


WIDTH:624
HEIGHT:352
DURATION:2545
FPS:23.976
VIDEO_FORMAT:XVID
AUDIO_FORMAT:mad
VIDEO_BITRATE:1007296
AUDIO_BITRATE:128000
AUDIO_CHANNELS:2
AUDIO_SAMPLING_RATE:48000


I separated the above for clearity, so don't think the script is doing it. But the order of the movies should be the same since i ran the commands from the same directory of movies. I didn't analyze the results, but will let you know later if i find something fishy. But the script is quite fast, and if thats all thats taking to get accurate results, then excellent job. idvid might end up being just as important as tovid itself.

Sweet. It looks like it's doing exactly what it's supposed to so far. For non-compliant video, maybe I'll add a line like:

Audio is compliant with the following video disc formats:
None

Just to make it more clear what it's looking for. I'm glad to hear it ran fast for you; I did have a problem video (a video file with no audio stream) where the duration-estimating part took as long as the video itself was, which would be very bad for a 1-2 hour video; I might put some kind of timeout in there, so if it hasn't figured out a duration after 3-5 minutes, it stops and uses the duration reported by ffmpeg or mplayer -identify instead.

Sounds good, the "Audio compliant" line does look empty without have something like "None."

So, is this output going to become parts of the 'stats' file? It looks really cool, so i would put it in there.
Let me know if you want more test, i have a few other avi's, but am out of mpg's.

Also, i am a little lost as to how 'idvid' is doing the above...is mplayer playing the movie to find the length and then some other program is giving the info?

I wasn't planning on putting the output in the stats file, since I'm sort of assuming that you'd be using idvid for finding out whether your video is (S)VCD/DVD compliant, or to get other information about it. I'll probably also start using it from tovid, to have a more accurate duration, and to skip encoding stuff that is already valid. Just email me more tests if you find something wrong, or if you identify other mpegs. I'm especially interested in seeing how well it identifies stuff you've encoded with tovid, to make sure tovid is encoding things correctly.

Actually, mplayer/mencoder is doing all the identification now. The duration comes from mencoder, which runs through the video really fast (hopefully) to find out the ending time. The rest is from mplayer -identify, with some minor cleanup and correction of its output.