Crash if trying to rotate interlaced video

What is your operating system?
Windows 10 x64

What is your Shotcut version (see Help > About Shotcut)?
23.05.14

Can you repeat the problem? If so, what are the steps?

-load an interlaced video in shotcut (* I can also see the crash even if I just first manually choose interlaced in the properties panel of a progressive video)
-open the properties panel and change the orientation to 90 degrees (assuming it was at 0)
-100% chance to crash for me

This doesn’t crash if using an older (old = qt5) version: 22.12.21.

Thanks for your report. I reproduce this. I will add it to my TODO list.

For a workaround, apply the rotation with the Size, Position and Rotate filter instead of the clip properties.

1 Like

A few more curious details:
-the videos [I was loading up] are supposed to be progressive but both shotcut and mediainfo reports them as interleaved - is there a sure way to confirm which is the real scan mode? These are supposed to be 1080p25 shot on a sony camera that doesn’t have 25i mode (but has 50i)
-what happens if I manually switch to interlaced mode a video that is progressive (and otherwise)? I don’t actually notice any change (other than the fact that I once had a crash after changing the settings too often)

-also vaguely helpful to reproduce faster: if you save a project with a video in playlist with scan mode interlaced + rotation 90 degrees, the next time you reload that project file it is enough to just double click the playlist item to produce the crash

Edit: forgot to finish the ideea of using progressive mode on an interlaced detected clip: if I do this I can just rotate with no crash without using SPR, so I’ve been using this as a workaround

In the Shotcut properties panel, there is an option for “More Information” that can give some more clues.
image

But this can all be confusing and difficult to interpret. For example, the container might signal that the stream is progressive, but individual frames might signal as interlaced - or the other way around. This difficulty in interpretation is the reason we allow the user to override the detection.

25i and 50i are the same thing - only one is indicating the frame rate and the other is indicating the field rate. And manufacturers decide not to standardize and always use one or the other - so we just have to know that no matter which they write, they always mean 25 frames per second / 50 fields per second.

The only change is that you might trigger Shotcut to use a deinterlacer if the source is set as interlaced and the output is progressive. Else, you might not be able to see any difference.

That should be fine if your clips don’t have any combing artifacts that are common in interlaced sources. Be sure to look at the exported output to make sure there are no interlacing artifacts which are hidden by the default “one field” deinterlace mode in the shotcut preview player.

Oh man, you were right about this being confusing. I ended up going into a deep rabbit hole for an hour or so. It seems my camera wrongly marks the file as interlaced top field first (as reported by multiple tools) but using the idet ffmpeg filter it seems 100% of analyzed frames are progressive (and also tested with an actual 60i recording from camera and it did report 100% interlaced).

So in conclusion none of this mattered because I exported a few tests and no matter how it’s marked in properties the output looks good and I guess this is what matters.

The crash is fixed for the next release.

1 Like