vspreview with vs_align

[jay123210599]

Like I mentioned before, I have different videos that I want to compare to, but they are out of sync. How do I use vs_align to perfectly align the videos in vspreview?

https://github.com/pifroggi/vs_align

[_Al_]

Again, vs_preview has nothing to do with your thread questions, it is just a previewer that previews whatever is set for output in you vapoursunth/ python script, you confuse yourself and others.

from vs_align, I did not install it, just looked in manual, after you install pytorch and numpy (for precision=1 you might not need that, not sure, it looks like frame stats are used only), you might use temporal alignment,
clip = vs_align.temporal(clip, ref)
where clip is misaligned clip and ref is "normal" clip (your reference clip), precision=1 is default. Those two videos should be almost identical, just misaligned for precision=1. If those two videos differ in quality and colors a lot, you'd need to choose precision 2 or even 3.
Then, at the end of your script, you set both outputs, clip and ref and compare them in your previewer selecting particular outputs.

[_Al_]

but for a simple clips misaligning of your clips you can use trim or slice feature and do it manually,

clip=core.lsmas.LWLibavSource(r"misaligned.mkv")
ref=core.lsmas.LWLibavSource(r"ref_version.mkv")
#guessing ref is 15 frames ahead of clip, so cut those 15 frames from ref's beginning
ref = ref[15:]
#guessing clip is 15 frames ahead of ref, so cut those 15 frames from clip's beginning
#clip = clip[15:]
clip.set_output()
ref.set_output(1)
so by error and trial you come up with fine value to match them

[jay123210599]

I have 5 videos/clips.

[poisondeathray]

I have 5 videos/clips.

Align each one to the reference

[jay123210599]

Align each one to the reference

Which video should be the reference?

[_Al_]

It does not matter. Your choice. Maybe the best quality. But it has to be explained that way, what is reference video and and what video is adjusted.

[poisondeathray]

I would choose that has the most similar shared frames to the others as reference

e.g if you have 5 versions, perhaps different edits, different cuts - pick the one that the most shared frames with all the others . If you pick one that has the most missing frames compared to the others, it will not be as good candidate for reference