Stabilize filter fails at video cut in timeline

What is your operating system?
laptop MSI cpu 10’ + geforce 2060

What is your Shotcut version (see Help > About Shotcut)? Is it 32-bit?
21.05.18 - 64bit

Can you repeat the problem? If so, what are the steps?
Stabilize filter add a file with X/Y directions as you can see editing the file.
I suppose that if I cut the video AFTER a stabilization file has been created, the match between the coordinates inside the stabilization file and the video fails.

As results, a perfect stabilized video, if cutted, moved and merged with a previous video in timeline, result in bad stabilization. Camera moves are present, zoom of stabilization are present, but they are not correct.

Moreover, as workaround, I could RE-ANALYZE the video-chunk BUT: analyze is frozen after first run.

So the correct workaround is to destroy the filter and add it again, repeating the ANALYZE.

So, I suppose, the Errors is a bat match between the STAB file coordinates and the video chunk. This occurs when you STAB, analyze, Cut the video chunck, move it, and create a transition with previous chunk.

I don’t know if without transition the error occurs…
I always use standard transition (cross-fading).

Can you check?
thank you


I agree this could be better.

In addition, if you split a clip that has been analyzed, the filter will be duplicated on both halves of the split and the same analysis file will be used on both filters.

If you change any of the filter settings, the Analyze button is re-activated. So you could change a setting and then change it back to reanalyze. Maybe we should change the button so that it is always active.

1 Like

We can do that for the next release since it is simple, and I agree to it as we already have the text feedback about filename and analysis.

As for the mismatching analysis data, we need to add code in various places to clear that property when a clip’s in point changes or split. Then, there is already code in export that will enqueue analysis jobs if needed.

1 Like

Hi and thank you for reply.
Stabilizer filter has a couple more of issues: when you set a stabilize filter to a video part, and you DO NOT analyze, but proceed to EXPORT, before exporting a message ask you to analyze the parts.

So there is a sort of internal check that verify that the analysis has been done before exporting the video.

The first problem is that this flag should return to OFF if you cut the video part, considering the previous analyze not effective. This should resolve bad stabilization problem (Of course, if can be found why it fails to synch video time vs stab time, is better…).

The second problem is that if you have video01, stabilized and cutted 5 times, but not analyzed, you get in your queue 5 analysis of the same video file. A huge amount of time and resources wasted. Stabilization in fact should be 1 to 1 whith video files. They could have the same name, such as video01.stab. And if this file already exists, the analysis can be skipped using this stab.

Just some tip to improve the process.
thank you.

It is not analyzing the entire file. It is only analyzing the trimmed portion. Doing the entire file once when you may not be using all of it will be slower. In any case, there is some significant re-engineering required to use multiple shots from a single stabilized source with some bad side effects unrelated to stabilization that I will not get into.

1 Like

I submitted this change for the 20.08 release.

I also did this for the 20.08 release. The change seems simple and low-risk.

I think there is more that we could do to make this better in the future, but I think these changes would be higher risk:

  • Nothing stops the user from selecting the same file name for multiple filter instances
  • We could show the name and location of the analysis file
  • The results could be cleared when the filter panel is not visible (due to filter panel being hidden, or a different filter being selected). We could find some way to draw the user’s attention to filters that need to be analyzed.