Help on DV to WMV

[DDogg]

Please move to another forum if this is not appropriate here.

Hello all, I could use some help if somebody is in the mood. I have a talking head guest lecture I need to encode to WMV and try to fit on one CDR-80 while keeping quality as high as possible.

Source is 80 minute DV type 2 (18 gig). I would normally just output the timeline in Vegas to WMV 9, but I thought it might be better to try to deinterlace and decimate to 23.976 to save a few bits. Maybe wrong in that, dunno.

First question is advice on script. Audio will sync ok, you think?
avisource("Way.avi")
InterlacedResize(544,400)
fielddeinterlace()
Decimate(cycle=5)
AssumeFPS(23.976,true)

Second question: What bitrate to use and what is the absolute simplest tool to encode to WMV. I am thinking 2 pass vbr would be best? I just don't ever do WMV from outside my NLE so my brain is revolting and since it is so damn long, testing a lot is going to make me miss a deadline. Any suggestions, advise, changes, or how tos would be appreciated. I just need to get this done.

/add I choose WMV because playback needs to be idiot proof and autodown any codecs needed.

[Neo Neko]

Originally posted by DDogg
Source is 80 minute DV type 2 (18 gig). I would normally just output the timeline in Vegas to WMV 9, but I thought it might be better to try to deinterlace and decimate to 23.976 to save a few bits. Maybe wrong in that, dunno.

Since we are talking DV then should it not be either true interlace or progressive if it is a high end DV recording? I am not familiar with any DV cameras that record at 24fps and then telecine. So decimation could result in any significant movement becoming jerky.

Originally posted by DDogg
First question is advice on script. Audio will sync ok, you think?
avisource("Way.avi")
InterlacedResize(544,400)
fielddeinterlace()
Decimate(cycle=5)
AssumeFPS(23.976,true)

Audio synch should be fine. Decimation should not change synch but changing frame rate can. But since after decimating every fifth frame of a 29.976fps video gives you a 23.976 video the AssumeFPS line should not effect synch. In fact it should not effect anything at all. I am not familiar with "interlacedresize". Where did you get it? Preserving proper interlace while resizing is impossible AFAIK. If you are going to resize you "need" to deinterlace fist AFAIK. If this video can safely be decimated of every fifth frame then shouldn't you use telecide instead of fielddeinterlace? And I am assuming this should be a 4:3 video or have a 1.333 aspect. So shouldn't you resize to 544x408?

I would suggest...
avisource("Way.avi").telecide.decimate(cycle=5).lanczosresize(544,408)

Originally posted by DDogg
Second question: What bitrate to use and what is the absolute simplest tool to encode to WMV. I am thinking 2 pass vbr would be best? I just don't ever do WMV from outside my NLE so my brain is revolting and since it is so damn long, testing a lot is going to make me miss a deadline. Any suggestions, advise, changes, or how tos would be appreciated. I just need to get this done.

/add I choose WMV because playback needs to be idiot proof and autodown any codecs needed.

A nominal BR of about 1Mbps should suffice if you are going to use 64Kbps WMA for audio since it is only voice. That should nearly fill an 80minute 700Mb CD. If the audio is Mono then you could go to 32Kbps and bump up the video bitrate a bit. You could even resample to 32khz instead of 44.1 and drop the audio BR more. As far as encoders there are only 3 others I am aware of. Windows Media Encoder and Nic's encoder which both take avs scripts. And Windows Movie Maker which I am not sure about. If this is pretty much total talking head video with no odd zooms or pans then you should be able to drop the BR for the video to 500Kbps or less and keep good quality. It all depends on what you wan't to do with the video.

[trbarry]

I am not familiar with "interlacedresize". Where did you get it?

InterlacedResize() is part of the SimpleResize() filter ( www.trbarry.com/SimpleResize.zip ).

But it should only be used when you have to keep to an interlaced format all the way through. While it does manage to resize interlaced material keeping to (I think) the correct sample points on the correct fields it will still not give the detail level of deinterlacing first and then using a standard progressive resize function.

And note InterlacedResize() is currently disabled for YV12 data in Avisynth 2.5 anyway. Sorry, I forgot I'd done that until this post. I'll try to fix that as soon as possible but there isn't much call for it and it just slipped through the cracks.

- Tom

[trbarry]

DDogg -

I'm also skeptical that your DV source can be smoothly decimated to 24.

I don't usually do this but if you really want to lower the frame rate for compressability then you might consider 20 fps instead of 24 as it would play at a more regular rate. The way to get this would be to use one of those TomsMoComp scripts I believe you already have to deinterlace to a full 60 fps, then select every 3rd record with SelectEvery(3,0). This should give a smooth progressive 20 fps clip that you could then resize and compress as needed.

- Tom

[DDogg]

Neo Neko and Tom, Thanks for your replies.

I was using Interlaced resize because of two reasons. I am a big fan of Tom's work and the darn thing is very, very fast. I always use it for resizing DV. Maybe it is not supposed to work properly if used before deinterlace, but then again, maybe my ignorance is bliss because it sure seems to work just dandy. Also, the script above seemed to be fine although I only did a few short bits to check ... seemed smooth...but then again it was just a simple talking head shot.

Tom, I am intrigued by your thoughts on using 20fps. SInce this is just for a disk handout for PC playing, that might be a real good way to go. I guess that would effectively increase the usable bandwidth by a third or thereabouts....hummm. Later.

[Neo Neko]

Reducing the frame rate will effectively increase the bits that will be allocated to each frame. But if this is all headshot and done from a stable tripod etc then I don't think video bitrate will be a real issue even if you encode to 640x480 29.97fps with 1Mb/s nominal 2-pass. The only thing that is generally easier to compress than headshots is a contiuous solid true black screen.

[DDogg]

Neo Neko, of course you are correct. Originally I was fixated upon the 1hr and 20 mins as I am a svcd'er. When I slowed down to think about it, I remembered you guys were getting incredible amounts of video on one disk and that the paltry 80 mins was nothing to worry about.

As I read the replies, I started thinking of the next "Honeydoo" (don't know if that translates) from my wife to do this as streaming video so Tom's remarks got my interest.

Thanks again for you and Tom taking the time to write your thoughtful replies. It was very helpful and got my thinking on the right track.

Best Regards,

DDogg

[Neo Neko]

Honey would you do this? Honey would you do that? It translates fine. ;) Though sometimes it is more of a demand than a question. hehe