Topaz Video Enhance AI - Chronos 25fps > 60fps video output is longer

Trying to go from 25fps to 60fps with Chronos with a 10 min 10s video, however the output comes back as 10 min 29s, tried with the original file (wmv) and also tried converting it to x264 in handbrake and running it again with the same results. The duration in the mediainfo is correct for the original file and shows the longer duration on the output file…

Any ideas? (Slow motion is set to 100%, FPS is set to 60)

(Had this happen with multiple videos too)

1 Like

When I contacted support, it seems that I am currently aware of this issue.
It seems to be a trouble that occurs when duplicate frames are included.
I hope it will be improved in the next chronos model.

1 Like

Hmm I’ll see if theres duplicate frames and trim them and see if it helps :smiley:

1 Like

Tried generating a new version with ffmpeg decimate to remove duplicate frames… seemed to remove some (dropped the framerate to 20fps), ran Chonos again and the output was even longer… so… no idea :smiley:

1 Like

The current version of the chronos slo-mo model doesn’t seem to work unless it’s 2x4x…
Probably due to 3x from 20fps to 60fps

Ah… that explains it, tried 25 to 50fps and it had the correct times. Hope it gets fixed soon!

I have created a procedure to solve this problem (23.976> 59.94 or 24> 60).

1 --Make the post-deinterlaced video (I’m using avisynth QTGMC) at 30fps with assumeFPS.

2 --Convert 30fps video to 120fps with Chronos slo-mo.

3 --Set 120fps video to 96fps with assumeFPS.

4 --Use avisynth “srestore” to thin out frames to frate = 60.
(In most cases, after this process, execute “ntsc_double” in assumeFPS)

This method is, frankly, very stupid, but currently there is no choice but to do it.
Please solve the problem of magnification other than 2x4x as soon as possible.:scream:

1 Like

In the current Slo-mo model, there are many blend frames because there is no scene change detection.
Therefore, flowflames RIFE is used to convert the original source to 60fps, and upscaling is performed with VEAI.