Prepping Two Source Videos With Different Frame Rates for Blu-ray

[RigganRose]

Hoping to get some advice with this issue. Had it posted on creativecow for a few days and no one would touch it...

I'm working on prepping a project for blu-ray using Sony Vegas Pro 13. My two source videos are different frame rates. The first and most important source vid is running at 23.976 fps. The second video's rate is 29.970 fps. Before I realized the two files were different rates I made a test bluray of the final cut to quality check for editing mistakes. This is when I noticed severe motion blur on my primary footage and realized the source video was running at 23.976 fps. I rendered a new video using 23.976 fps inserting 3-2 pulldown, and it has worked wonders in correcting all the motion blur that I saw with the larger video. However, as expected, the smaller source video that runs at 29.970 is now left with noticeable motion blur even after I set the video to "disable resample" before rendering my final file.

Is there a way to correct this in Vegas? I tried to convert the larger file up to 29.970 so the frame rates of both source vids would match. But the final result left a bit to be desired. Although the motion blur on the larger file improved, it is not as smooth as when I render using 23.976 fps inserting 3-2 pulldown.

The results on the larger file are where I want them. Just need to correct the motion blur of the smaller 29.970 video. Seems to me the way to go would be to reduce the secondary footage from 29.970 to 23.976 so when I render the whole thing using 3-2 pulldown both videos upscale smoothly together, but I'm not sure how to do that without that source video developing issues.

Opinions and plausible solutions would be appreciated. Thanks!

[poisondeathray]

SD or HD blu-ray ? Because HD blu-ray supports native 23.976p (no pulldown required)

Also, how are you determining this? What is you method of viewing? Hardware or software ? Did you physically burn a disc ?



The short explanation is vegas offers 2 types of framerate conversions, 1) blending 2) duplicating or dropping frames . Neither is very good, because most framerate conversions are not very ideal when they are not evenly divisible integer multiples (e.g. 59.94 and 29.97 for example, are perfect. Converting betwen 29.97 and 24 or 23.976 are not ideal). When you have frame blending enabled it will be blurry / ghosted images and jerky. If you disable frame blending (disable resampling in the properties) it will use the 2nd method and you will get jerky footage because every 5th frame will be dropped. If you disabled resample as you mentioned, and STILL getting blur - this suggests you are blend deinterlacing somewhere. That means your 29.97 source footage is probably interlaced and you have deinterlace set to blend. Set it to deinterlace: interpolate instead of blend in the project properties

The 3rd method that vegas doesn't offer natively is motion interpolation or optical flow. New "inbetween" frames are synthesized from motion vectors, so the frames are "evenly spaced" in time. The motion is the smoothest compared to the other 2 methods, but this method is prone to edge morphing artifacts. On some types of content it works very well , but on others it's very poor. You can get plugins like twixtor for vegas, or use free methods outside of vegas like avisynth

It's not possible on blu-ray to change framerates in the middle of a titleset (you can't mix 23.976 and 29.97 , they have to be the same). But you can use different rates on different titles

[RigganRose]

First off, thank you for your reply and good info.

To answer your questions:

The primary souce vid is HD. The secondary one is SD (VHS rip).

I did physically burn a disc. I have a re-writable blu-ray disc that I like to use to quality check my editing. I never like to trust the preview window in Vegas. After I reach the point where all my clips are in place and I'm 99% sure I'm finished, I like to burn a test disc and pop it in a home player to check for tiny flubs that Vegas might not show me. This is where I originally noticed the motion blur from the primary video. Now, after my 3-2 pulldown render the primary is smooth but the secondary one is now blurry and 'ghosting' during movement.

One thing I may not have been clear on in my initial post; these two source videos are getting rendering together as one video for the disc. They both are connected to one another and are blended together to make one file. I've done this successfully before with no issues (other than the obvious clarify difference between the two sources). But, I've not had the issue where the two source videos are running at different frame rates. This is where the issue lies with trying to get the secondary footage to look smooth after the primary source video gets the 3-2 pulldown treatment. But, I didn't know blu-ray could support 23.976. I've always been under the impression it had to fit the 29.970 standard or it wouldn't work properly. This changes things a bit. Should I render my final file at 23.976 instead of using 3-2 pulldown to bump it 29.970?

[LMotlow]

I don't think you're paying attention. Try reading poisondeathray's post more carefully. You can't mix two video types in the same clip without distorting one or the other in some way. It'll take fancy frame work in something like Avisynth to make it right.

I didn't know blu-ray could support 23.976. I've always been under the impression it had to fit the 29.970 standard or it wouldn't work properly.

You might want to familiarize yourself with the BluRay standard specs: https://www.videohelp.com/hd#tech.

Should I render my final file at 23.976 instead of using 3-2 pulldown to bump it 29.970?

That still won't fly.

[kitaro]

I have same kind of problem with my 3D camera : it records at full HD at 25 fps. But the 3D blu ray specs says 24 fps. So the video was choopy : every 25 frames, one was dropped to make it 24. So my actual solution is :
- render the project at 25 fps (in my case in powerdirector) in M2TS file
- split audio and video
- change frame rate of the video to 24. The tool I use (MKVMERGE) does just change in the header that the video is 24 fps : IT DOES NOT RE-ENCODE the video
- stretch the audio to slow it down a 1/25 slower so that the length match the new length of the video.
- merge new audio and video
--> I have now a 24 fps.

Of course, the whole video is slower and thus longer (by a 1/25th), but it is nearly not visible and of course better than dropping 1 frame on 25.

In your case, you should do the process to your source files. But problably moving from 23.976 fps to 29.976 fps may be more visible than from 25 to 24.

It worth a try.

Good luck.

[LMotlow]

That won't work. The 29.97 will play too fast.

I just now took a 29.97 interlaced video clip and decoded it into lossless interlaced avi. Then I took a 23.97 progressive clip and encoded it to a high bitrate mpeg with 3:2 pulldown to make it 29.97fps. I then opened that telecined mpeg in VirtualDub and saved it as lossless 29.97 avi, without removing the pulldown. I then imported both 29.97 clips into an encoder, and the encoder imported both of them as "interlaced" 29.97 video. I encoded those joined clips together as 29.97 interlaced. Worked OK for me, and I didn't see any combing on the formerly telecined frames. Will take another look at it later and see if it looked as correct as I think it does.

I did all that work in Avisynth and VirtualDub before sending it to the editor/encoder.

[RigganRose]

I am aware the final file needs to be the same frame rate. The whole point of this post is figuring out how to make that happen.

I feel the best solution to this issue would be to convert my smaller, second priority video from 29.970 fps to 23.976, as the main source video looks best when left at 23.976 so I don't want to change the frame rate on that one. What is the best way to make that happen? I'm not familiar with Avisynth other than some light reading I did when this problem first occurred. Never actually used it. I am familiar with VirtualDub. Which one would be the best in handling the conversion?

[poisondeathray]

That's what most people would do, use native progressive for HD blu-ray, when your (main) source film rate 24.0p or 23.976p. The resulting encoding quality will be higher too, than if you used 3:2 hard pulldown (interlaced encoding, 25% duplicate fields - you can think of it as wasted bandwidth)

But there is no "perfect" solution to your framerate problem as mentioned earlier. They all have problems. You have 3 basic approaches mentioned above. Vegas can do the 1st two. Definitely method 3 will give smoother results, and the artifact severity depends on the source video characteristics - it sometimes works well, sometimes it doesn't. You can try it out with mflowfps in avisynth since it's free, or maybe a trial version of twixtor for vegas. Vdub has MSU FRC as a plugin you can try too. There are more advanced conversions that involve layers, masks , motion tracking to improve the motion estimation and quality of results - but that is much more complicated you probably don't want to get into that right now. But if you choose to try method 3, and your 29.97 source is interlaced, you should bob deinterlace to 59.94p before doing the framerate conversion. Deinterlacing is another huge topic...

[LMotlow]

You've got it backwards. You can convert 23.976 to 29.97 by adding 3:2 pulldown, them saving the video as decoded/interlaced AVI. The reason you save the telecined video as decoded AVI is because decoded AVI has no telecine flags or 3:2 pulldown markers. Then both videos in decoded AVI form xan be imported into most NLE's/encoders as "interlaced".

You can convert 29.97 video to 23.976 video by blending and/or dropping some frames or fields, losing quality, losing clean motion, always seeing blend ghosts and/or combing/blurring, and getting some really weird looking video. Didn't you complain about that in post #1?

See post #6. I just did what you propose, but in reverse. What's wrong with telecine or interlace? You watch both on TV all the time and when you buy DVD and BluRays.

[poisondeathray]

Yes, that's the other option - make an interlaced blu-ray, where the "23.976p" sections have 3:2 hard pulldown applied so are encoded interlaced, and the 29.97 sections are obviously interlaced - so the BD is running at 1920x1080 29.97 interlaced. But it has potential issues as well. If you ignore the lower quality / compression efficiency considerations - playback problems will depend on the hardware setup . That alone is usually enough to give people pause about taking that route. Cadence reading players can adapt and IVTC the "23.976" sections properly, and bob deinterlace the 29.97i sections properly as well. So everything is "perfect" in theory. The problem occurs when you don't have a HW setup that can handle it properly and adaptively swap. That sounds like what the OP has since his physical disc playback didn't handle either test very well. ie. The sections can be mistreated - e.g. the 23.976 sections might be deinterlaced instead of IVTCed giving you half resolution, jerky footage , or combination of "fails"

[LMotlow]

Yep.There vare some players that never really do anything. They have so much latency they just sit there.

The alternative is to load both videos into the timeline and get the same results as before.

That just about leaves this thread with no place else to go. Finally.

[RigganRose]

That's what most people would do, use native progressive for HD blu-ray, when your (main) source film rate 24.0p or 23.976p. The resulting encoding quality will be higher too, than if you used 3:2 hard pulldown (interlaced encoding, 25% duplicate fields - you can think of it as wasted bandwidth)

But there is no "perfect" solution to your framerate problem as mentioned earlier. They all have problems. You have 3 basic approaches mentioned above. Vegas can do the 1st two. Definitely method 3 will give smoother results, and the artifact severity depends on the source video characteristics - it sometimes works well, sometimes it doesn't. You can try it out with mflowfps in avisynth since it's free, or maybe a trial version of twixtor for vegas. Vdub has MSU FRC as a plugin you can try too. There are more advanced conversions that involve layers, masks , motion tracking to improve the motion estimation and quality of results - but that is much more complicated you probably don't want to get into that right now. But if you choose to try method 3, and your 29.97 source is interlaced, you should bob deinterlace to 59.94p before doing the framerate conversion. Deinterlacing is another huge topic...

Gotcha. Thanks for all the helpful info.

I'll definitely test and see what happens with the Vegas methods you've given, but seems Avisynth might be the way to go here. I'm a little gunshy since I've never used it and it's command line interface is a bit quizzical to me. I've seen a couple GUI's available that promise to make Avisynth easier to use. Is there one in particular that you'd recommend?

[poisondeathray]

There aren't really any good GUI's for avisynth

There is a bit of a learning curve, and a bit of a hassle to collect the proper .dlls and scripts, but once you get going it's really simple . It's not really commandline usage (not cmd.exe), rather it's script based. You can probably just copy & paste a script, and don't worry about what is going on (you can spend time learning about the basics later when you have time)

Start with posting some basic info on your SD footage. Use mediainfo (view=>text and paste back the info here). Do some small tests before you commit completely (recall I said some types of footage characteristics don't respond well to optical flow ) . Did you want to convert it all or just select portions ? Think and plan ahead of what you're trying to do to save time converting something you're not even going to use. Also It's not easy to replace your edits (you can use the replace function in vegas, but I don't think it will work with different FPS clips, so you 're probably going to have to massage the project a bit. Also at least one of your your previous projects was at 29.97, so you'd have to redo one as 23.976p anyways)

It's critical that you bob deinterlace the SD footage as a first step (I'm still making the assumption that it's interlaced SD, correct me if I'm wrong but it's usually pretty safe assumption when you have SD at 29.97) . That will give you 59.94p to start with because 29.97*2 = 59.94 where every field essential becomes a frame. You don't have to read this next part but what you're really doing is upsampling to 119.88fps by interpolating every 2nd frame 59.94*2 = 119.88 (synthesizing an "in-between" frame), then taking every 5th frame to give you 23.976 (119.88/5 = 23.976). That's why they are "evenly spaced" in time and motion is smoother compared to the other methods

[RigganRose]

I see. That's cool. I'll give it a whirl and see what I can come up with.

Oh, and to clarify; here are the specs on the SD footage I'm needing to convert:

Type: Video for Windows
Size: 20.27 GB (20,753,373,696 bytes)
Created: Wednesday, July 01, 2015, 7:02:02 PM
Modified: Wednesday, July 01, 2015, 8:32:47 PM
Accessed: Sunday, July 12, 2015, 4:14:14 PM
Attributes: Archive

Streams
Video: 01:30:44.272, 29.970 fps interlaced, 720x480x24, DV
Audio: 01:30:44.300, 48,000 Hz, 16 Bit, Stereo, Uncompressed

[poisondeathray]

It uses DV compression, but was it sourced from a DV camera ? 99% of native DV is interlaced, but there are 24p, 24pA variants which already have pulldown (e.g. some professional DV cameras shoot this for "theatrical" shoots) - can you clarify if you have the "normal" variety with interlaced content ?

[RigganRose]

No, I don't believe it was sourced from a DV camera. It was shot with a VHS camcorder in the 80's. The DV is probably coming from the USB DVD recorder I used to feed the footage into my computer for editing.

[poisondeathray]

VHS camcorder means it's almost certain to have interlaced content

An avs script is really a text file. To the recieving application that opens up avs scripts, it looks like a "video". You can use notepad for example, and change the extension from .txt to .avs. Usually avspmod or vdub are used to preview scripts.

1) First thing is install avisynth, 32bit. Stick with the vanilla , non multithreaded , 32bit version for now. Most filters and plugins and dependencies are 32bit for the avisynth stable branch. Once you get comfortable with avisynth you might want to check out the MT branch, or even vaporsynth. Like everything, there are pros/cons to each, but the stable avisynth branch is the most robust and most commonly used.

2) Next gather the dependencies, .dll's. Most .dll's are autoloading when placed in the avisynth/plugins folder. Or you can load them in the script with LoadPlugin() .

I recommend QTGMC for general use deinterlacing. It is far far better than what vegas offers. You can do a quick search if you want to see some comparisons, but really there is a big difference. I typically use it in one of the faster settings for general use. You get 95% of the quality but it's much faster. There are dozens of different settings that you can read about in the documentation at your leisure. You can also get about a 2-3x speed boost using MT, but it can give unstable results - I'll leave that for another discussion

In this thread there are links to collected filter dependencies as a single download package, so you don't have to spend hours searching all over for them
http://forum.doom9.org/showthread.php?t=156028

A template for your script is going to look something like this. Again, there are dozens of different variations, but this will get you started. You have to replace the PATH with the actual path, and filenames (e.g. "C:\folder\yourvideo.avi")

AVISource("PATH\video.avi")
AssumeBFF()
QTGMC(preset="faster")

source=last
super = source.MSuper(pel=2)
backward_vec = MAnalyse(super, overlap=4, isb = true, search=3)
forward_vec = MAnalyse(super, overlap=4, isb = false, search=3)
source.MFlowFps(super, backward_vec, forward_vec, blend=false, num=24000, den=1001)

So you would edit the path & filename, save it, rename the extension from .txt to .avs (e.g. script.txt to script.avs or whatever name). You can preview it in vdub or avspmod. Scrub through and check the motion, also look for problem areas. Report back if something seems out of the ordinary. You can playback avs scripts in x86 MPCHC, but you probably won't get realtime playback on this particular script (too slow). You can save a video to import into vegas with vdub or avspmod . Typically what is used is a lossless codec like ut video codec, huffyuv, lagarith, or some near lossless codec like cineform. You can also do edits with avisynth, or with vdub, or avspmod (e.g trim to wanted sections)

I probably missed a few things, but hopefully that will get you started. If you get stuck, usually avisynth will post an error message that indicates what's wrong when you preview the script, so ask here with the message and someone will help

[RigganRose]

Thank you very much. You've gone above and beyond what I was hoping for when I came here. I appreciate good people like you. I'll try all three methods you suggested and see where things are. Many thanks, Buddy.