VFR to CFR at Exact Rate: Why does correction quality vary so drastically

[Migue12HD]

Hello community.

I'm working on converting a large batch of VFR (Variable Frame Rate) screencast videos to CFR (Constant Frame Rate) using high-precision methods (VfrToCfr or MKVToolnix) with the timecodes.txt file. My goal is maximum fidelity with zero interpolation.

I've observed an unusual pattern and seek technical confirmation:

Hypothesis: Is it normal for the VFR to CFR correction at the exact average rate to yield radically different results (perfect fluidity vs. intense stutter) depending on the inherent instability of the source video?

Example:

One VFR video at 56 FPS average is corrected to CFR at 56 FPS and results in perfect fluidity (VFR stutter eliminated).

Another VFR video at 67.312 FPS average is corrected to CFR at 67.312 FPS (exact fraction) and the result is stutter that is worse than the original VFR video.

My deduction is that the 67.312 FPS video had such an unstable frame cadence (corrupt microsalts in the timing) that when the VfrToCfr filter is forced to apply the CFR correction using that defective timing information, it introduces abrupt and noticeable jumps.

Is this 'corruption of the camera/recorder's cadence' the reason why the timecode correction can fail so spectacularly, regardless of the mathematical precision of the target CFR rate?

Thank you in advance for any insight.

[rgr]

In the second video, some frames are probably recorded at, for example, 20 fps, so these frames will produce a less smooth, jerky image. Without transition frames, the image won't be smooth. You can send me a short clip to test.

Besides, you don't need any plugins or scripts to do what you want. Just load it into VirtualDub and adjust the fps value if necessary -- in this program, each frame is treated as a CFR.

Another problem is that with this type of video processing, the sound will be out of sync with the image.

You may also be interested in the TimeCodeFPS plugin (AviSynth).

[poisondeathray]

Converting VFR to CFR by inserting duplicates , or dropping frames, won't make it any smoother than the original.

Usually if it looks jerky, the original was jerky to begin with . ie. Low FPS sections . The MIN fps in sections is low. Inserting duplicates in that case won't make it any worse - it will be as bad as the original

But in some cases, the VFR to CFR conversion can be worse than the original when you DROP good frames to achieve the desired FPS . Those cases generally have a higher MAX fps in sections. For example if you had 80 or 90 actual FPS in a section with unique frames, and you are dropping good frames to make 67.312 FPS CFR, that section is going to be worse than the original.

Also, the refresh rate matters - eg. if you are using a 60Hz display, you cannot "see" 67.312 FPS . The display is functionally dropping frames

[rgr]

Also, the refresh rate matters - eg. if you are using a 60Hz display, you cannot "see" 67.312 FPS . The display is functionally dropping frames

As VRR becomes more common, this doesn't matter as much.

[poisondeathray]

Also, the refresh rate matters - eg. if you are using a 60Hz display, you cannot "see" 67.312 FPS . The display is functionally dropping frames

As VRR becomes more common, this doesn't matter as much.

VFR has already been common for many years - eg. phone videos

Refresh rate frame drops are independent of VFR , and it matters whenever smoothness of a video is being discussed. Whenever current frame rate > refresh rate , then you have frames that are not displayed (dropped)

eg. Pretend you have smooth 90/1 FPS CFR true gameplay video. You view on a 60 Hz display. You cannot display all 90 frames per second, when the refresh rate is 60. Frames are dropped by the display and you get jerky playback. There is additionally a judder cadence, because the frame drops are not evenly spaced apart in time

Now if you had 120FPS true CFR on that same 60Hz display, because 60 is evenly divisible into 120, the judder is absent because you display every second frame, but it's still less smooth than displaying 120FPS on a 120Hz panel