Wanted Frames Removed

[MrVideo]

The video is MPEG-2 4:2:2 ~57 Mbps

I took two seconds of black from another portion of the program, then the program and added the same two seconds of black at the end.

The two seconds of black is at the beginning of the file, but VRD dropped 1009 frames of the program before starting to include the program. The two seconds of black are at the end.

Why?

[MrVideo]

I tried doing a small section to see if I could start the program at a different location, only to ultimately go back into an area that is liked even worse, i.e., the YTOD appeared.

Am I not allowed to work with 4:2:2 video and create TS files?

I was able to output ES files, but I do not know if the output was correct, because of the issues here.

UPDATE: Nope, doing an ES results in the loss of the 1009 frames that I want.

[MrVideo]

I went back farther in time to a chapter marked area of the video section I'll be keeping for a different part of this project.

That worked. It started at exactly where I wanted it and went through all of the video, without errors, including the part that it refuses to start from, without getting the YTOD.

[Dan203]

4:2:2 is only supported if you have the multi-audio setting set to 0. This is because in that old stream class we use a special encoder, different then the one used for the rest of VRD, to recode the few frames around each edit point. That special encoder supports 4:2:2, the main encoder that we use for everything else does not. Just like with H.264 we have a "pro" version of the main encoder that does support 4:2:2, but it's more expensive to license so you'll have to talk to DanR if you're interested. (or you can switch back to the single audio stream option)

Dan

[MrVideo]

4:2:2 is only supported if you have the multi-audio setting set to 0. This is because in that old stream class we use a special encoder, different then the one used for the rest of VRD, to recode the few frames around each edit point. That special encoder supports 4:2:2, the main encoder that we use for everything else does not. Just like with H.264 we have a "pro" version of the main encoder that does support 4:2:2, but it's more expensive to license so you'll have to talk to DanR if you're interested. (or you can switch back to the single audio stream option)

So, why can't this encoder be used for multi-audio as well?

This is a one-off project, as I don't normally get stuff that is 4:2:2, but this feed was :mad:

There are seven, yes seven, audio streams that go along with the feed. Of which I need to keep them all. It would mean seven passes to get audio stream output and with seven sections to edit, that is a lot of passes.

Since I'll never recode the edited sections to 4:2:2, the need for the expensive encoder doesn't exist. The final output will be H.264 for Blu-ray.

[Dan203]

The MPEG-2 multi-audio class uses a new modular design, similar to the H.264 class, which can only use a single encoder for both quick edits and full recode output. Unfortunately the encoder that supports 4:2:2 is really slow, so it's not practical to use for full recodes. So we had to go with the other one which only supports 4:2:0.

4:2:2 is really uncommon in the consumer space, so for most people this is not an issue. And like I said we have a special "pro" version of our encoder that does support 4:2:2 for our commercial clients, but it's more expensive to license so we can't include it in the consumer version of VRD.

Dan

[MrVideo]

And like I said we have a special "pro" version of our encoder that does support 4:2:2 for our commercial clients, but it's more expensive to license so we can't include it in the consumer version of VRD.

Unfortunately, to buy a more expensive version, just for a single project run, is not in my economic interest.

Guess I'll be busy :D

[Dan203]

If you set the multi-audio to 1 and then output using the "Audio Only" profile it will output all of the audio streams to separate ES files while ignoring the video stream for syncing. Perhaps you can combine these with the 4:2:2 video output with the single stream class using tsMuxer and save yourself some time?

Dan

[MrVideo]

If you set the multi-audio to 1 and then output using the "Audio Only" profile it will output all of the audio streams to separate ES files while ignoring the video stream for syncing. Perhaps you can combine these with the 4:2:2 video output with the single stream class using tsMuxer and save yourself some time?

Are you saying that by using the audio only output the fact that it is 4:2:2 video will be ignored and actually cut the audio as per what I have in the joiner? Then I go back and do it again with audio set to 0, but this time output both video and audio?

That will save some time.

[Dan203]

I believe so, but you'll have to try it. The "Audio Only" option takes a path similar to recoding the whole video, rather then using the quick edit logic, but then just throws out the video frames as they come into the graph rather then recoding them. I think this will allow it to bypass the issue with the 4:2:2 video. But I've never tried it so I can't say for sure that it will work.

Dan

[MrVideo]

There is an issue with changing audio support, specifically going to no multi-audio from multi-audio. It is not honored within the current session. Then, if you restart VRD and read in the project file, the setting in the project file overrides the config setting.

[MrVideo]

OK, that trick appears to be working. I have to make two project files, one with the multiaudio on and on with multiaudio off. Simple file copy and editing takes care of that.

[MrVideo]

I have to retract that it worked. It failed when going through the joiner. The edited piece should have been 44m7s (44m3s program and two 2s blacks). When going through the joiner, with multiaudio on, audio only output, the length of the audio pieces is 43m33s.

For the helluv it, I tried something. Because it appears that using the joiner screws things up, multiaudio or not, I just did a Save As on the edited piece, multiaudio and it worked. The video started on a B-frame and VRD still encoded the beginning frames, not losing a single frame. Joining, not so good.

I don't have time to try it at the moment, but the next thing is to create a 2-sec black 4:2:2 TS file and join it with the edited piece (head and tail) and see if that works, since no recoding should take place.

[Dan203]

I think you're going to run into the same issue. VRD uses the same quick edit coder for output regardless of whether recoding is actually required or not. So if you attempt to join the 2-sec of black to the file you're going to have to use the single stream class and you'll lose your extra audio streams.

The only way this would work is if you created a 2-sec black file that had the exact same number of audio streams, with the same specs, as the source file. (must contain actual audio frames) Then joined that to the original file using the "Audio Only" output mode. The "Audio Only" output mode will only output audio frames in the source file, it will not create blank audio frames to fill the gap like normal joiner output does.

Dan

[MrVideo]

While you were posting, a test run was being done, which has now finished.

First off, the black was extracted from the end of the feed, so it is technically the same.

I put the separately created, multiaudio, black into the joiner, twice (once for the head and once for the tail). Then I read in the edited video and placed it into the joiner, in between the two blacks and ran the job.

Perfect output, no loss of frames.

So, individually, 4:2:2 output can be created and initial frames recoded, without frame loss. Said individual files can then be joined, without issue.

But, try and use video that has the black, the program video you want to edit and put it all together in the joiner, it tends to fail.

The thing I have not tried yet, and will shortly, is to take the predited black and join it with edited project video in the joiner and see if that works.

[MrVideo]

The test run is now complete. That also worked. By using premade 2 sec black files, along with video that is being edited, the result is what is expected.

I have no idea why that works and using all edited project video doesn't, is beyond me.

[Dan203]

I don't know either. I'm having a hard time even following what exactly you're doing. But I'm glad you got it working.

Dan

[MrVideo]

Obviously editing 4:2:2 video. It will ultimately be recoded with x264 for ultimate authoring onto Blu-ray. For that reason, there needs to be at least two seconds of black added to the head and tail of the edited video. You do not want to start, or end, DVD/Blu-ray video on program material.

So, the segment of video that will go onto the Blu-ray will consist of two seconds of black, then program video, then the two seconds of tail black.

The black piece comes from the program itself. When the feed was done, they went to black. So, I created a project file containing two seconds of black.

When the black project file is placed into the joiner (twice), along with the program project file, VRD ended up dumping many seconds worth of program video at the joint of the header black and the program (I think the tail joint was bad as well). Has to do with the way you handle 4:2:2 video.

Then with your suggestion about trying to output elementary streams in no multi-audio and with multi-audio, resulting in some bad results as well, I discovered than by just doing a Save As on the program material, multi-audio, TS output, that there were no frames lost.

Ultimately I discovered that by creating a file containing just two seconds of black, it could be added to the head and tail of project program video without a loss of video frames.

I have no idea why pre-made black works when using a project file of the same black video does not.

It is not worth trying to track down further as I'm probably the only one that uses 4:2:2 video in a consumer environment.

[pendragon]

It is not worth trying to track down further as I'm probably the only one that uses 4:2:2 video in a consumer environment.

It may seem that way, but there are a number of other FTAers like me that have to deal with this also. Generally I try to keep my high-rate 4:2:2 files in their original pristine condition on file servers, and minimize their exposure to VRD for some of the very same reasons discussed here. I often wish I didn't have to jump through so many hoops to perform simple edits on these outlier formats.

[MrVideo]

Glad to know that I am not the only consumer that sometimes has 4:2:2 MPEG-2 video.

[DanR]

I've got tons of 4:2:2 files here (from our broadcast users). The last couple of builds of TVS4 has much more 4:2:2 support during output including resizing, 4:2:0 conversion, etc. If we missed something we would like to know about it.

Can you detail how to duplicate?