An automated Vegas 24p workflow..
 

We've been discussing in another thread here about automating the 24p workflow. Bill Fransworth and Mike Dulay came up with a a workflow using DGIndex and VirtualDubMod and some scripting to put it together, though there were still a few manual steps.

The thread and what needs to be installed is here: http://www.dvinfo.net/conf/showthread.php?t=92893

I put together a Vegas script DLL that further automates the process so you really don't have to do a whole lot of anything to remove the pulldown. The script will ask for a working directory (should be a clean directory that you don't mind the script to delete or copy things to) and the paths to the DGIndex and VirtualDubMod applications as well as the AVI Synth template. You can specify the script to parse the required audio delay as well, which it will gather from the AVS file generated by DGIndex. Another option is it can create a new 24p vegas project and put the output from VirtualDub into the timeline (so you can re-render to your choice of codec etc..).

Next is the compression used by Virtual Dub. I put a couple templates up there for Cineform, Lagrinth, Uncompressed AVI etc. Just because they are there doesn't mean you have them installed though, so make sure to pick one you like. If there is another one you'd prefer, you can choose the "custom" option and create the internals for Virtual Dub job script yourself. You just need to leave out the script calls for VirtualDub .Open, .subset, SaveAVI and Close. My script will add those automatically with the correct frame settings and file names.

The final option (second tab of the dialog) is the list of media clips found on the selected video track. Check the checkbox next to those that you want to run the script against. I strongly recommend just one track and one video clip for now though.. It shouldn't screw anything up otherwise, but hey, this is kind of a beta here..

So here is the script (my first one and a bit of overkill personally):
http://www.szudzik.net/vegas/scripts/24pWorkflow.html

Click save and place it in your vegas scripting folder.

I'm still going to work on this some more. Right now, it will process the whole file represented by each event on the timeline. It ignores clips, markers and all of that. I just grab the VideoEvent and get the underlying media file path and will operate against the entire file. As a result, it's best to just test this with a single track and a single video on the track.

The script dialog also allows you to load and save "template" scripts that you can use to update the script dialog for how it operates. This doesn't apply to the list of clips, but only to the general settings tab.

The "Clean" button on the dialog will DELETE ALL files in the working directory specified on the dialog. Make sure that's what you really want. I'm not responsible if you delete anything accidentally! Really, I'm not!

The Close button works as a "cancel" or "done", click it when you want the dialog to go away.

The "Run" button will start the workflow process. Depending on what options you choose and how big your file is, it might take a few seconds or significantly longer. Each of the resulting AVI's generated by VirtualDub will be placed in the working directory with the same name as the original video file, but with _24p.avi appended to the end.

I think that pretty much covers it for anyone who wants to give this a try. Like I said, try 1 clip first and see if it does what you want. I'm going to get a page crafted with the above info and some screen shots to serve as my "Help" file, hopefully done in a little while with that.

I'm going to comment the source code and will later make that available as well for anyone who wants to make changes or improve the workflow.

Again, this is "BETA" type software here, so it's a use at your own risk type of thing. I'll fix up any issues you guy's find and listen to suggestions for it as well though.

--Steve

Ok, I found one small bug (new I rushed it a touch) and have posted a quick updated DLL on the page. The bug caused the process to fail the first time through because it didn't detect the AVS generated by DGIndex. The second time through worked properly. It now works the first time as well as subsequent times.

--Steve

Some notes
 

Steve, you may want to update your instructions on how to invoke the script dll file. I was looking around Program Files when I realized you could store it anywhere. Ran into a roadblock with the Cineform HDLow default compression in the script.

How I ran:
1) Downloaded Vegas trial from sony
2) Installed and registered trial version
3) Dragged and dropped an m2t file unto the Vegas timeline (setup for HD size with 23.796fps)
4) Stored your szudzik.HV20Pulldown.dll unto c:\clip (arbitrary choice -- probably should be C:\Program Files\Sony\Shared Plug-Ins)
5) Invoked Tools -> Scripting -> Run Script... -> specify location of the dll
6) Specified c:\temp as my work directory
7) Specified C:\clip\24p\dgindex\DGIndex.exe as my dgindex path
8) Specified C:\clip\24p\VirtualDubMod_1_5_10_2_All_inclusive\VirtualDubMod.exe as my virtualdub path
9) Specified C:\clip\farnsworth.avs (this Farnsworth's original template with -222ms ... from here ... http://www.hv20.com/showpost.php?p=1421&postcount=10)
9) Selected Calculate delay audio
10) Left the default option CineformLowHD
11) Run

DGIndex runs okay, generates all the files in c:\temp, syl file is generated and virtualdubmod 1.5 launches. Now I get the immediate message "License Required: This application is not licensed for CineForm Hd encoding. Please visit www.CineForm.com to purchase your upgrade." I click okay and virtualdubmod continues to run. The generated avi file in c:\temp displays the cineform logo (too bad, the file looked good and was only 1x larger vs 33x). When this is dragged on to the vegas timeline you can see that the logo is hardcoded into the avi. I think you should rather default to Uncompressed (but recommend in the dialog LagarithLossless or HuffYUV codec).

I repeated 10-11 but with Lagarith this time.

12) I add back the 24p AVI file to the timeline ... rendered to HDV intermediate. I can probably use this as my archival format and edit format (but not as a playback) ... for final render I would use a common codec like AVC/H264 or WMV.

One thing I can comment, it is so clean from a use perspective! I think you'll help sell more copies of Vegas. 8-)

Thanks for the feedback Mike. Yeah, I just threw those instructions together quickly. I'll get them updated soon hopefully. I today also got the dialog from VDub about the cineform codec, guess my trial has ended ; ). It still won't uninstall but interestingly, I don't have the logo either. Maybe a bug with vista and how they uninstalled.

For everyone else. I have also made this "script" into a stand alone executable that you can use to automate and batch the workflow. You can find it at the below location:

http://www.szudzik.net/vegas/scripts/HV20Pulldown.html

It's C#, so requires the .Net Framework 2.0, just like vegas but his doesn't have any Vegas dependancies. The UI is nearly identical to the vegas script, so this page should look familiar.

The difference here is that one the 2nd tab, you can select all the media files, from wherever you want and add them for processing. This version also does not require the copying of the file to the "working directory" so it's much faster.

Let me know what y'all think..

Steve

Downloaded your exe file. Will report how it works once I get some 24p footage this weekend! ;-)
Thanks for all your effort and time!

No problem, hopefully it will work well for you. I might make a few more updates to the standalone app this weekend. I might "hide" the DGIndex window. It's such a short process, so you don't gain much by watching it work. I wouldn't mind hiding the VDub window as well, but since I can't get any progress info back from it, I'm reluctant to do that. At least with that window up, you can see what's going on.

I think I prefer the standalone app right now myself. The Vegas one can be good, but I'd have to put in a lot more work. It would be nice if it honored the splits and markers etc.. Someday maybe ; )


--Steve

Working like a champ
 

Steve, your standalone exe is working great. I've got a harddrive full of m2ts that I'm converting right now. Wish my new laptop was faster but hey, I'll take what I can get. I'll have to make a new blog referencing how to put your exe together with all the other freeware software. Can't have you doing all the work. 8-) Thanks for doing this!

Thanks Mike, glad that it's working well for you.

I just updated my documentation for the application at: http://www.szudzik.net/vegas/scripts/HV20Pulldown.html

Put in the links to the required apps as well as a better workflow on how to get started. I also made some small changes to the app last night. Got rid of the "clean, run, load, save" etc.. buttons and put them in a file menu. Looks a bit cleaner. Also doc'ed most of the code and should be posting that up shortly.

Funny thing is there's more code for working with the UI than there is just to run the workflow! Could easily be fully scripted in a batch file as well, except then it's tougher to select the video files, unless you just batch process an entire directory.

--Steve

What I might be doing wrong?
 

I downloaded all the scripts and the self-executable file. Also had to install xvid. The program works and gives an output. But the output definitely doesnot look to be progressive!! (While playing I can see the combing/ghosting).
To test I did the following to compare....
1) Converted the m2t file to avi.
2) Opened the avi in VirtualDub.
3) Used the telecide filter
4) Played the output-- this definitely does not have those artifacts and is progressive. Looked at it frame by frame. But it does have one duplicate frame (every 5th) as it should.

So what I might be doing wrong???

Odd ...
 

Ajit, are you manually doing the process rather than letting Steve's exe create the avi for you? The exe should have done the TFM and TDecimate for you already. #3 on your list should not be necessary as there is no longer any pulldown for you to process with telecide.

Hi Mike,
I tried doing it using Steve's exe. But that did not work out! That's what was my point saying that the output is not progressive!
However when I used the filters manually in virtualDub it gave me progressive.
can you think of any possible reason?

Ajit, Steve's exe isn't standalone. You probably did not install the TIVTC plugin into avisynth. Look for post #26 on this thread ...

http://www.dvinfo.net/conf/showthrea...t=91500&page=2

I did understand that Mike. I was following the whole development from Farnsworth first post at the other forum.

Anyway I might have missed installing something. I am going to check the whole thing again. I do remember that I downloaded the filter but I have to check if I installed it properly.

I'll get back soon.
Thanks..

If the specific VirtualDub script that you selected in my app (I call them "compressor") doesn't quite do what your looking for, you can use VDub to get the settings how you'd like. Do that and the click "Save AVI" but instead of saving, tell it to create a job. You can then look at the jobs in VDub and save those to a file. Open the job file and you can then bring that in as a "custom" script on my UI. Just leave out the Open, Close, SaveAVI and .subset portions of the script.

It could also be something in the .AVS template file as well. I know that Farnsworth & Mike both have a few different settings between them.

Anything you can achieve by just using DGInex & VDub can be done in my app. I'll I'm doing is automating the files through those two applications.

--Steve

Can you tell me when I install the TIVTC plug-in, I just have to copy the TIVTC.dll file into the avisynth plug-in folder. Am I correct?

Yes, that is correct.

--Steve

Worked!!
 

Hi Steve,
I made your script work for me finally! Well the problem was not with my setup but I was doing something stupid while viewing the footage. Now the footage does look progressive. Thanks for all the help guys!
Well still there seems to be some issue with the audio. When I open the avs file in virtualDub it seems a little off. ( It says a framerate of 24.12 would match the audio perfectly instead of 23.976). Is that what you are getting?
And Farnsworth modified his script to include a delayaudio for 222 ms. Do I need to keep that while using your script? (which also includes another delayaudio command.)

Suggesting Modifications
 

Hi steve,
can I suggest the following modifications to your script if doable?

1) Using VirtualDub instead of VirtualDubMod. As the latter one is no more developed it seems. The issue seems a minor one as the avs files from the script does work with VirtualDub if done manually but it gives an error if used via automated script instead of VirtualDubMod.

2) Instead of defining the codec's within the program if the program could list all the available codecs in a particular computer where its installed.

3) Once the script run is finished to generate the .avi file it would give an option to delet all the .mpa, .d2v and .avs files, so the output would be just the avi file.


I hope I am not asking for a lot!
Thanks

Ajit B.

#1 VirtualDub would be just fine, it depends on which line of the script is producing the error though. I can play with that and see if I can make it compatable.

#2 would be great, I totally agree. I would love to pop up the same dialogs that VirtualDub does for picking which compression you want. That seems to dive into using the Video for Windows API though. I'd like to get that, but will probably take me some time since I'm not very familiar with that API.

#3. Yup, was thinking about that as well. That request is very simple to perform. I might get that one done today.

As far as the audio delay is concerned, if you check the box on my app, what I'll do is check the .AVS template produced by DGIndex (in the working directory) and pull the ms delay it's specifying for the audio delay. I'll convert that and update the AVS with a call for DelayAudio with the appropriate value. Like Mike has mentioned in some of the other threads, each clip needs a different offset. .222 was just the one that happened to work for Farnsworth for his particular clip.

You can also tell my app not to try to do the DelayAudio and then you can put your own value into your Template.AVS file, which will just get passed along through DGIndex..

--Steve

Good morning Steve!

Well the error that comes up as a popup window while using VirtualDub says "Error during script execution at line 1, column 30: Member 'RemoveInputStreams' not found
VirtualDub.RemoveInputStreams<!>(); "

and as I told if you open the avs file manually in virtualDub later there seems no error. I hope this will help you to figure that out!

2) For the API call is it possible to look through the VirtualDub source-code? That might make the things easier for you! (May be a stupid idea! as I am not familiar with programming in windows!)


Well can you tell me the source of this audio delay. I mean how its getting created? Then perhaps things would be clearer to me.

Thanks

Ajit B.

Yes, I have the error here as well. The reason you can open the AVS file (and I can as well) is that the REmoveInputStreams call is actually in the VirtualDub script file that I create (32Pulldown.syl) in the working directory. That's the file I pass to VirtualDub. If that line is not present, then even VirtualDub is happy. I'm not sure that's a good thing though...

I've found two other web pages referring to that error. Unfortunately one is in Spanish and the other appears to be in Russian. Google was able to translate the Spanish one, but there was no resolution that I saw there. I'll keep digging into it and see if I can figure that one out.

Yes, I downloaded the VDub source and had already done what you were suggesting. It's C code, which I'm a bit rusty at but I was able to get the gist of what they were doing. I just need to figure out how to convert that into C# code, two very different beasties for sure. I've seen some other C# code using VFW, but none around this particular set of calls though. Once I figure that out, we'll be golden.

Honestly, I'm not 100% sure what DGIndex is doing to figure out the delay. Mike might have an idea (he was the one that pointed me to getting the delay from the DGIndex generated AVS file).

Oh, I've got the request you made for cleanup done. I'll be posting that in a little while. Just trying to see if I can also get the VDub issue addressed as well.

--Steve

Great Steve! At least some of things are progressing!
About the audio I think this delay is generated by DGIndex itself. So I am thinking if there are some alternative freeware to use which could do the same job.
May be MEPGstreamclip or something like that. What do you think?

I've posted my updated binary to address the "post cleanup" actions. It's a checkbox next to the Delay Audio option. I also remove the "Video Files" tab and put it all together on one page.

Yes, DGIndex is determining that delay value. I have seen a few other threads talking about using MPEGStreamClip, but have not looked into them too much, so couldn't say how that would fit into the workflow.

--Steve

Avoiding DGIndex
 

Steve,
It seems we can avoid DGIndex by just open the m2t file directly!! Though the audio is still not perfectly in sync but its much better. (VDub with Farnsworth method and correcting audio sync with your script tells to put video to 24.15 instead of 23.976 to time them exactly.. while this method I am using is saying 23.955).

Well the way to open the m2t file directly the command is

DirectShowSource("xxx.m2t")

My avs script at the moment is

DirectShowSource("capture01-0064.m2t")
AssumeTFF()
telecide(guide=1, post=2, vthresh=30)
decimate()


and this works!!

Edit:
I am using the decomb filter from neuron2.net instead of the TIVTC plugin. They are almost same.

And thanks for your update! I am going to download it now.

Ajit,

MPEG Streamclip was the first method I tried. It did a deinterlace rather than a proper 3:2 pulldown removal. On occasion you get odd blips and blocks in the re-encoded video.

What OS are you on and which codecs do you have installed? Before using Farnsworth's method I had also tried the DirectShowSource method but VirtualDub/Dubmod would keep crashing when I dragged the AVS into it. I'm not sure if this is a Vista problem or my choice of codec packages (my main came with CCCP and SUPER(C)).

That variable frame rate you're getting is strange. Mine has always come up 23.976 whether using Farnsworth's or my version of the avs. Maybe its DGIndex, could you post a sample of your generated files if it persists?

Steve,

DGIndex examines the MPEG transport stream (think of TCP/IP packets and OOS reassembly) to generate the audio delay. It's a function of how the HDV codec laid down the video and audio from the camcorder. Since the start/pause and cadence isn't always the same per clip each one has a different delay. M2T files are close kin to TS files and M2V files you find in HDTV transmissions and DVDs. DGIndex source code is open, the routine for reading the stream should be in there, but would it be worth it to reinvent the wheel? Maybe if we were doing it directly from HDVSplit or our own capture program ... hmmm ...

I too have been getting a consistant 23.976 fps in all of my test clips.

I did get the DirectShowSource to work, I'm also using Vista as well. Been using it for over a year at work and couldn't stand not having it at home! I don't think that Vista would be the problem for you Mike, but you never can tell.

I had thought about capturing myself, but I just don't have that kind of time to write the app! Guess that's why the Cineform folks get the $$$ ;).

--Steve

Well Mike now I kind of understood where this audio delay is getting generated. I had the idea that DGIndex is "creating" this audio delay so was thinking of replacing it with someother program that might put audio in better sync.

And abut the frame rate I think my explanation was not clear enough. I do get 23.976 fps. Those strange frame rates are the one required to match the audio and video durations exactly (as reported by VirtualDub from Video->frame rate) and neither Farnsworth script or Steve's exe can put the audio in exact sync in any of my clips. (I hope I am clear this time!)
But I think I should not worry about this small duration issues.

Oh and my os is win XP2 sp2. I have Lagarith, MSU and huffyuv codecs installed apart from some Intel ones that I think came with something! (I didnot install them).
I did install vista after building this desktop. But it needed toomuch resource to run itself and many a softwares I use werenot supported so I rolled back!

Ajit,

Since you're not demuxing the audio track its still trying to play itself at the original length. DGIndex is the most efficient way I know to get the delay.

http://avisynth.org/VariableFrameRateVideo

There is an avisynth script that interpolates the delay which isn't as efficient:

http://avisynth.org/FindAudioSyncScript

I thought when I am using steve's exe (instead of my script) with the "Calculate required Delayaudio" box checked in it should demux it and give me same duration of video/audio. But thats what is not happening ( evident from those strange framerate requirement to match timing).
Or did I miss something?

Ajit,
When DGIndex runs it demuxes the audio track into an mpa (MPEG Audio file). The filename contains the delay value and that is what we (Steve's exe and my original PERL) read and recode into the Delayaudio() line of the avs script.

Anyway I think I should tell you guys that I am very new to this kind of videoprocessing. My knowledge of video processing was just regular cuts, color and gamma corrections (inherited from my still photography) and transitions.
And I learnt a lot in this few days from you guys and also through reading.
Thanks a lot!

well thats precisely what I think is going on in my case too. I do have the mpa files created (when using steve's exe) and that also included one delayAudio term in the avs script files. (different values for different file).

So I thought when I open this avs scriptfile in VirtualDub/DubMod I should find the duration of video and audio are exactly the same. But that is not happening. Can you check that with one of your avs files?

Ajit, I'm quite new to this as well. I've done some basic video modifications in the past with my old analog cam & pinnacle studio. I've only been in the "digital" world for a few months myself and boy o' boy was there a wealth of stuff to learn! Ok, I could still just stick with the basics and not worry about all of that stuff, but I'm a geek at heart and am trying to soak up as much as I can. Working on little projects like this helps because it forces me to start digging into these apps to see how they tick..

--Steve

same here ;-)

But my problem are more as I stopped coding about 7 years ago since my job does not require coding anymore! But I do work with scripts in linux/unix sometime.
So I find I need to read a lot more when it comes to windows!

I haven't done any C or C++ in a while so I'm pretty rust at that. We switched over to using C# at my job about 6 or 7 years ago and I just haven't had the need to use C++ since then. I can work my way through the VDub C code fairly well, so I at least know what's going on. Translating that all back into C#, that would take me a bit longer which is why I haven't done the compressor dialog yet!

Linux & Unix, yeah I'd be totally out of my element there! No opportunity to work with those at my job, they kind of dislike the philosophy from what I can tell ; )

--Steve

BTW, my cable modem appears to be on the fritz, so if you have a problem downloading the application, try again later. Service tech is supposed to be out on Wed. I'll put it up at an alternate site tomorrow though...

Steve

LOL, we're a bunch doctors trying to fix an engine. :-) My programming skills are rarely used as well.

I've checked my audio vs video and their lengths are not exactly the same. 47.23 seconds of m2t vs 46.26 seconds of demuxed mpa. Could it be a measurement error due to MPEG-Audio compression? I had clips that are several minutes long but the talkies were in synch for my ears. Somehow they've stayed in synch via audiodub(). It's probably smart enough to do the temporal adjustments when you change the framerate.

What if you try something like this:

DirectShowSource("capture01-0064.m2t", fps=29.97, convertfps=true) # or fps=23.976

There are some notes about problems with DirectShowSource and synch:
http://forum.digital-digest.com/showthread.php?p=530584
http://www.videohelp.com/forum/archi...e-t323397.html

Hmm.. you are probably right mike. This little difference might be due to compression. And in my case also I am finding them to be completely in sync to my ear!
But I have only converted small clips. So cannot tell what would happen if you have a longer (say about 10 min) clips. I would presume if its due to compression the differences will go smaller and smaller with larger clips. If not then there is something to worry about.
I am wondering can we forcefully speed up audio a little bit to match the video. Is that possible?
I am afraid I cannot devote time in video learning before the weekend!

I understand that! Well I personally like the GNU philosophy and most of the software tools I use in my work only run in linux/unix! So could not have helped it even if I disliked! ;-)

Ok, I've published a new version of the standalone application. Updated binary and instructions are at:

http://www.szudzik.net/vegas/scripts/HV20Pulldown.html

What changed:
Not too much. I made the compressor templates more customizable so that you can create a new one (or delete existing ones) and they will get saved to disk to be re-used later. This pretty much only affects the Script Editor window and the "compressor" drop down list on the main dialog.

--Steve