mask.py get_mask seems to be broken about lvsfunc

[LightArrowsEXE]

Thanks. Can you elaborate on the "broken" part in the title? As far as I understood it, if the mask clip doesn't match the given clips in MaskedMerge, it will simply use the last frame of the mask clip for the remaining frames. Does the old code actually break anything relevant?

[Selur]

Used the mask_d = lvs.mask.BoundingBox((0, 785), (665, 294)).get_mask(clip) and ended with a clip that is one frame long instead of the length of the clip (~2hr @23.97fps).
Problem there is:
a. the documentation states something different
b. when I use:

mask_d = lvs.mask.BoundingBox((0, 785), (665, 294)).get_mask(clip )
mask_d = core.std.InvertMask(mask_d)
mask_e = jvs.flat_mask(clip , radius=5, thr=5, use_gauss=False)
mask_m = core.std.Expr([mask_d, mask_e], 'x y +')
shrp = core.warp.AWarpSharp2(clip, thresh=200, blur=2, depth=32, planes=[1, 2])
filteredClip = core.minideen.MiniDeen( shrp, radius=[4, 1, 1], threshold=[10, 10, 10])
clip = core.std.MaskedMerge(clip , filteredClip, mask_m)

Problem here is that mask_m only is one frame long which is the bounding box merged with the first frame of clip and that mask is applied to the first the whole clip, which is in no way what is expexted. So yes, at least for me it breaks stuff. (+ the documentation if wrong) Maybe it works if mask_d and mask_e were switched in the core.std.Expr-call,... haven't tried.

[LightArrowsEXE]

Gotcha. Well, now it properly returns the full-length clip, so that shouldn't be an issue anymore. If there's any other issues left, message me in the discord channel or leave another issue 👍