In the Export queue, the status had something new and odd - it said “Running task 1/35”, instead of time remaining. The tasks continued, 2/35, 3/35, etc.
When it finished, the Status said “Scene concatenation failed” and there were 36 small mov files in the directory that it never combined back into one.
I’m on a MacMini M2.
I’ve closed the app and am running it again to see if I get a different result, and will update when I know. I have the logsforsupport file ready.
Looks like the particular file in the logs you should look at is “2025-07-18-10-41-57-Main.tzlog”
Possible clue: For the first time ever I clicked on the “scene detection” option in the Framerate Conversion panel - and each mov fragment appears to be an individual camera shot. So I am going to try unchecking that box as a third try.
Another clue: The short mov files are numbered, but the sequence is not perfect.
It starts with “_s1” but the next is “_s4” then “_s8” then “_s10” (where did 2, 3, 5, 6, 7, and 9 go?) and only at _s10 does the numbering continue onwards one at a time.
And that was consistent on both the first try and the second try (missing 2, 3, 5, 6, 7, and 9).
There is a description of how the new scene detection workflow runs and a few pieces of information on what can happen in the event of an error.
From what you described it sounds as if the workflow is working as designed but could be running into issues with short cuts in the video source.
The logs do contain an error with the file path of the input source on the last attempt as it is being read as not existing. Can you check that for the input source and ensure it did not get moved or altered after being loaded into the app?
The input source was not moved or renamed since it was loaded, but I can run some more tests later on this and other videos. I may try simplifying the file path too.
I see in the release notes that shots under a few seconds may error out (“…have at least 8 frames per cut within your video”), and that I should have seen (but did not) a “Retry Failed Scenes” button.
Music videos tend to have pretty quick cuts, and in this particular music video they used a sort of cheesy stop-motion effect which likely was interpreted as a series of very quick cuts.
So I assume that caused the problem, and ultimately prevented the segments from being “stitched back into a single video”.
As Scene Detection is refined, this problem will likely go away.
I am thankful that you made Scene Detection though. Thank you for adding this much-requested element.
I’d say mark this as resolved, since although it is not fixed per se it appears to just be part of the “growing pains” of the new Scene Detection option.
Likely you are correct that the model is having a hard time with the stop motion effect in the video and this will be something the research team has to work on when it comes to artistic to creatively edited videos.
