Stabilization analysis not triggered by export if stabilization filter copied from another clip that already was analysed

Bug
#1

What is your operating system?
Linux (latest ubuntu) 64bits

What is your Shotcut version (see Help > About Shotcut)? Is it 32-bit?
22.06.23, should be 64bit AFAIK

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

Haven’t found a way to reliably repeat this in small test project. For big “real” projects it seems to happen quite often

When one source clip is split to multiple pieces and stabilization filter used, and mixed with other clips, exporting video from timeline causes alert for analysis job not ran for all pieces, and asks if I want to do that. Then if permission granted it progresses to run the same analysis over for every piece of the source file.

As far as I can tell, the analysis is for whole source clip, and just once per source should be enough (after running it once the filter settings screen says analyzed for all pieces). For simple test project running analysis just once per source file is sufficient.

→ the “analyzed” status is not always propagated to all pieces using the aready analyzed source clip, despite the filter settings screen showing it.

#2

When you analyze a segment of a clip, it only analyzes that segment, not the entire source clip.

As an example…

  1. Open a clip
  2. Put the clip on the timeline
  3. Split the clip in half
  4. Add the stabilization filter on the first clip
  5. Click Analyze

This process only analyzes the first half of the clip. If you also want stabilization on the second clip, you need to add a filter and then analyze that part of the clip. Additionally, if you trim the in or out points of a clip that has already been analyzed, it needs to be analyzed again.

This might not meet your expectation, but it is by design.

If you have a large clip that requires stabilization and you also plan to cut it into segments, I would suggest to open the clip, apply a stabilization filter to the entire clip, and export the stabilized clip to an intermediate lossless file to use in your project.

#3

Not true. It analyzes all of the source clip (at least the current stable version). And for most times no need to rerun it.

Just like in tutorial (2:30 on) Shotcut - Tutorial part 35 - Stabilise - YouTube

It’s just a bit inconsistent when handling large projects. When trying with small test project one analysis run per source clip is, as I already mentioned in the opening, all that is needed.

For me the expected behavior definitely would be just analysis of the part where filter is applied, but for reason or other it does not go like that. My guess is that splitting it would be complicated.

#4

I am the developer who wrote the code for the stabilize filter. It works the way I describe, not the way described in that tutorial video.

As an experiment, you can do this:

  1. Open a clip on the timeline
  2. Split the clip at 90% so that the left clip is 90% and the right clip is 10% of the clip
  3. Add stabilize filter to both clips
  4. Click “Analyze” for both clips and give different names for the .stab file
  5. When the analysis completes for both clips, compare the .stab files.

The file for the left clip will be considerably larger than the .stab file for the right clip

This is the reason the analysis must be re-performed for any changes to the clip.

#5

Then something weird is definitely going on.

1 insert a clip to timeline
2 split it to 4 parts
3 put stabilize filter onto all parts
4 click analyse on one

Result: all show analysis done and for most cases do NOT ask to analyse the others when exporting

#6

Also I haven’t been asked for name of analysis file, they just seem to be assigned numbers.

#7

Aha!

Now I found out what actually happens (unless its still more complicated):

when copy-pasting filters from clip to clip, and when the source of copying has been analysed, that status is copied along the filters (at least to when pasted to part of same source clip). And no analysis for the other parts of the same source clip is triggered, and export is happily run without analysis and without warning.

This also explains why the stabilization seems poor sometimes… no actual stabilization happened as material was not really analysed despite filter settings screen saying so.

So whether filter is added “individually” to the clip or copy-pasted (along with color grading etc) the results are different.

image
image1847×1044 248 KB

For this biggest one analysis was never run but still says “Analysis complete”. Just coypasted the filter settings from the smallest. Asking to export also does not trigger analysis.

In what situation it should ask for a name for the stab file?

1 Like
#8

Updated the topic to reflect the actual problem.

1 Like