What "free" 24p workflow for Windows users?
 

I'm playing around with converting my 24p footage to 24 frame progressive avi/mov/mp4 file with free tools. Why not Cineform NEO? We'll its not free. 8-) The choices I'm aware of are Avisynth and MPEG streamclip. At the moment I'm using the latter. Would like to hear from anyone trying either tool and what settings you are using.

I would appreciate info on this topic as well.

Plus, I'd like more info on using After Effects to remove pulldown. (FYI, there's a free trial version, so it qualifies...at least for 30 days.)

found a previous thread on this same topic ...

http://www.dvinfo.net/conf/showthrea...t=24p+avisynth

Currently trying the following workflow:

1)HDVSplit -- extract video
2)HDTV2MPG -- convert to MPG for use with virtualdubmod
3)VirtualDubMod -- framerate 23.976, Deinterlace - Adaptive

Tried various compression, apparently WMV9 appears best with XviD following. All of them still had the some occasional interlace lines but it was least pronounced on those two. Uncompressed RGB produced stuttering video with lots of interlace lines (about 3.5GB vs a 150MB M2T file). Maybe I'm not doing it right. Will post source and samples as I get them hosted.

UPLOADED FILES:

Source m2t http://www.sendspace.com/file/oy3n90
Xvid output http://www.sendspace.com/file/u2ii3p
Wmv9 output http://www.sendspace.com/file/nlmg9u
DivX output http://www.sendspace.com/file/9de0ag

.. still uploading ...

in the meantime, I tried mixing a project together in a 23.976 timeline with both 60i and PF24 recording. When it was spit out this is what I got.

http://www.youtube.com/watch?v=Ch9pAx_c_qs

The youtube version is small but even at 1440x1080 the compression didn't show the horizontal interlace artifacts. So it makes me wonder if its worth it to do pulldown before editing instead of doing it in the final product.

I did a test "HDV 24p edit as 60i" in Vegas 7 the other day (will do more soon). Going out to a SD DVD (mpeg-2 60i) I couldn't see anything that looks different than the "true" 24p DVD's I've been doing with my DVX100 the past few years (though it took a million times longer to render HDV to SD DVD than my DVX DV files do.) However, I did a render of the same clip to a WM9 720p 24 in Vegas today and I saw some ghosting on movement in the scene. Not sure what that's about.

For me, the big advantage to doing pulldown (automatically) before/while editing is, 20% fewer frames to render (23.97 v. 29.97). This translates into faster render times and more room for video on a DVD. But I think most of us can edit our 24p files as 60i and not notice a difference in IQ.

Still, I think what we should do is contact Canon and ask for a firmware update that inserts flags for automatic pulldown removal in our 24p files.

Peter,

That's a great idea. How about a petition? Do you think anyone of influence at Canon would be reading this?

Well I can't say anyone of influence at Canon is reading. But I can say we users have much more influence than we think.

Started a new thread about this: http://www.dvinfo.net/conf/showthrea...150#post671150

I know you're looking for a free solution, but give NEO HDV a try, it is free for 15-days, and you might like the single step low hassle workflow for 24p.

Any idea when NEO will work for Windows Vista? I'd love to "kick the tires" on it, but it won't install properly (the regsvr32 calls all crash). Of course the app just crashes when trying to capture since the registration of the controls failed during setup..

Steve

We do intend to add Vista support shortly.

Great to hear. I'm really looking forward to trying out NEO. Since I've at least partially had the trial running now, will I be able to re-install the trial when you guy's do get Vista support, or will it timeout since it'll be beyond the 15 days?

--Steve

Unsure, we will have to wait for the update and see.

Found a method put together by Farnsworth using avisynth.

General flow:

1) Extract from tape via HDVSplit to m2t files
2) Run go.bat which calls DGIndex to detect cadence, output files to d2v(index of cadence in m2t)+mpa(audio) and generate a Avisynth file to do pulldown removal
3) Open avs file as the video source in VirtualDubMod. Framerate is already set to 23.976, all you need to do is specify the compression. Save file to a highly compressed AVI (compressed, but good motion reference) for your edit version, save using YUV or Lagrinth (each frame is a JPEG--"nicest" version but won't natively playback nice) for final render source. These are your true 24p source.
4) Open the edit AVI in your choice of NLE for editing. Then replace the file with the render version to get the best output.
5) When all done, you can delete the big YUV/Lagrinth (you can always re-render from the m2t later)

This style appears to work well and was fairly quicker than using MPEG Streamclip and without the interlace artifacts. One thing to note about the instructions, you have to copy over the .dll files from the plugins zip to the Program Files\Avisynth\plugins directory or you can't open the files in VirtualDubMod.

Also, for output formats I'm using Lagrinth lossless (big files! 1.5GB vs 150MB) for the final source render and WMV9 (9MB vs 150MB) for the edit source. (A technique of bait-and-switch). It's no cineform but it does a good job.

Here's a sample:
http://www.sendspace.com/file/bb4rt4


What's the difference vs the earlier method? Its amazing what difference you get in look from 3:2 pulldown (get back the progressive frame) vs adaptive deinterlacing (~guess and reconstruct what a frame should look like).

Having some audio sync difficulties with the above method. Simply adding the mpa to the timeline results in a variance that grows during playback. I was doing it via the NLE which is probably the mistake. If I can figure out how to remux in virtualdub or avisynth this may work well.

Another idea, if DGindex can detect the cadence, I wonder if it's possible to insert the pulldown flag into the m2t filestream. I don't know much about the format, but if it's possible that would save a lot of trouble. Then again, someone else would have come up with that plan earlier.

LOL, the idea to force insert of pulldown flag on the m2t wasn't going to work. m2t is MPEG2 Transport format, so it's video and audio encapsulated into transport packets ... oh why does my hobby have so much in common with my job ... long story short,a blind byte/word replacement wasn't going to work. The next best option would be to get the HDVSplit source or ask the author for an exit so the 24p could be reassembled there before encoding. That or you can throw money at the problem and buy Cineform (not an option for me as I use Vista).

In the meantime, I still go with the poorman's option (ironic, this hobby is expensive 8-). Determination pays off and I've figured out how to get the audio to sync properly.

Here's a sample:

vid=MPEG2Source("d:\capture\Auntshirley\\AuntShirley.m2t_.d2v")
aud=MPASource("AuntShirley.m2t_ MPA PID 814 DELAY -237ms.mpa").delayaudio(-0.237)
audiodub(vid,aud)
TFM(d2v="d:\capture\Auntshirley\\AuntShirley.m2t_.d2v")
tdecimate()

The next step is to automate it all. Farnsworth's windows batch file won't be able to parse the filename. We'll need to use PERL for that. Unfortunately it may be overwhelming for non-tech heads. We can simplify it as we go along.

Oh another tip, the MSU Lossless codec makes the smallest! And for previews the XviD codec has a highdef profile.