Usecase - a screencast recording where one is filling online details and scroll a lot arond…need to blur sensitive data that just keeps moving along with scroll action. Also, the sensitve data occupies the multiple parts of the online form (multiple text boxes of form) at random places in form.
From what I can see from this tutorial you have to go frame by frame to change where the blur effect resides. From my understanding, Shotcut would need to adopt a machine learning approach with object recognition to follow number plate / license plates. I’m excited by that idea but the size and editor’s OS may require multiple approaches, increase the overall exe’s size, etc.
It is not necessary to put keyframes on each frame of the video. The video I posted above is about 15 seconds long at 24 frames per seconds. So a total of 360 frames. But I used only between 15 and 20 keyframes per licence plate.
Here’s how I proceed most of the time:
I add a keyframe on the first frame and position the blur effect
I go to the last frame, add another keyframe and position the blur effect.
I go back to the first frame and play the video
I’ll add keyframes when the blur effect starts to deviate from the target.