| Wilbur Jaywright (zargulthewizard) | |
| Mar 8 2022, 8:57 PM |
| Peter Fog (tintwotin) |
| Pratik Borhade (PratikPB2123) |
| Richard Antalik (ISS) |
| Wilbur Jaywright (zargulthewizard) |
System Information
Operating system: Windows 11 Home 64 bit
Graphics card: NVidia GeForce 2060 RTX SUPER
Blender Version
Broken: 3.0.1
Exact steps for others to reproduce the error
Check resolution of separated strips - they are all same, but they should not be.
Interestingly, some images seemed to scale correctly, but others did not. These were all generated with Python PIL. I've marked where the time cursor was when I ran "Scale to fit," and images that remain the wrong scale. I hope this helps.
Note: I added these images by forst adding them all as a sequence, then separating them with Y.
Hi, still I'm unable to reproduce. Did you try with factory prefs?
Can you confirm the steps shown in the video are correct? (Doing same in your file)
Also, check again with latest 3.1 release- https://www.blender.org/download/
I did reset factory preferences. Note that this is Blender from the Microsoft Store. Here’s the steps from a default Video Editing file:
Move the time cursor around. Some images will go outside the preview frame.
I will test the latest build.
Actually nevermind, I realized, that problem could be in separate images steps and indeed, it does copy resolution of first image to all separated strips.
This is by design, because if this was 10 minute render, it would have to read 14000 images, which would take some time.
I think that this deserves a better solution. There have been requests to use same system for image sequences as with other editors, but there are some challenges to be resolved. This would mean, that VSE would be able to distinguish between sequences and batch of discrete images. However if you want to have sequence with variable resolution, I am affraid VSE can't support this scenario very well.
So will confirm this as known issue for now. If I did not describe workflow that covers your case, you can provide more info so I can refine this a bit more.
I checked if it would be better to set resolution of split strips to 0, so scale to fit would skip these images, but during rendering each rendered image will be corrected. So I won't do that, because this may cause problems for users that use sequences with uniform resolution. But current workaround is to scrub over images, to ensure data is correct, then apply scale.
Just FIY, I added edit with workaround to comment above.
Current operator could be changed, either:
In any case, if you want to have a go at this, feel free. I quickly checked what would take to do this some time ago and in short, image sequence filename parsing is in different module, so this code would have to be made public so VSE can use it. Then it constructs filename on demand and then tries to load it - VSE works in reverse, which is slight problem (need to parse filenames first and then handle exceptions). Or if only single image would be used as basis for sequence (as other editors do it), then you would have to iterate over all files anyway to determine length.
So it's not impossible or even super hard to do second option, but there is no great solution. It would be decision between ugh and meh, which is never great to review.
The Image Editor already has a Detect Image Sequences function, so it would be consistent to add one to the VSE import options too(downside is that the VSE import options are unavailable in the workspace embedded File Browser):
NB. the selected files are the posters of the Blender movies and not an image sequence - they will always be imported as a sequence in the VSE - which they aren't.
