Best method when rendering single frames from a video file.

[e.korpse]

Hello again.
So for the past couple of months I've been extracting images/frames from a whole video file using ffmpeg and then put them back together into a video using Blender. Then I extract the timecodes from the orginal video and mux them with the video that I rendered using Blender so I can get the video back to the right FPS. To summarize it:
1/ Extract Frames from video file using ffmpeg
2/ Edit frames
3/ Use Blender to make an image sequence to video using the edited frames.
4/ Use MKVCleaver to extract the video and timecodes since MVKCleaver wants me to extract the timecodes with a video track.
5/ Extract audio file using MKVCleaver
6/ Mux video, audio, timecodes together to get a new video using MKVTOOLNIXGUI.

I extract the frames at the FPS that original file is in. Now most of the videos I work with seem to be in VFR (Variable Frame Rate) and not in CFR (Constant Frame Rate).
I found converting the video to CFR using Handbrake make things easier because there were times when I tried working with a VFR, the audio and video would not sync and the video's FPS would increase and would also increase the time of the video. With CFR I have not come across to problems.
My other problem is that when I extracted the frames from the a video file I had the issue that not the right amount of frames would be extracted. Lets say I used Mediainfo and it says the video is 45000 frames long and when I use ffmpeg it would extract more or less than that. And when I tried doing all those steps I posted above I had the problem of the video and audio not syncing.

So my question is, is there a RIGHT WAY of extracting the frames from a video file without later coming to video audio syncing problems?

[Bernix]

If is it CFR and you know exactly frame rate, than it is no problem. Just tryed in Shotcut and it export exact number of frames. But somebody has problem with this software. When you started it you have to wait a moment (if you want to try it). But problem is you have to set correct fps.
I try it on video from my camera CFR 25 and it worked perfect.
And you can then put them together in virtual dub.
I would blame Handbreak for VFR videos, because it has it as default and it is almost imposible to work with such video propertly
I think the number that shows Mediainfo and ffmpeg capture, depends on framerate in your ffmpeg script and framerate of video.

Bernix

[e.korpse]

Thank you Bernix for your response.

So what your saying is that I should always work with CFR? And why is it so hard when working with a VFR? I know MKVCleaver supposedly has a VFR detection option. I'm not sure what it does though, I only know it extracts two timecode files.

[aedipuss]

vfr sucks. it should only be used as a last resort for saving a microscopic amount of file size. in other words only by apple crap.

[Bernix]

I think, that VFR is problem, why you get more or less frames. In CFR there is not problem, at least from my short experience.

Bernix

[jagabo]

why is it so hard when working with a VFR?

Because you have to keep track of time codes. Once you've lost them there is no way to tell how long each frame is to be displayed. CFR is much simpler, every frame is displayed for the same amount of time.

[e.korpse]

Thanks guys for your opinions.