- Kaunas, Lithuania
- https://aras-p.info/
- Joined on
2022-01-20 -
-
@hamza-el-barmaki this feels certainly out of scope; implementing whole "word art" is perhaps best left for some far future when VSE integrates way better with the rest of node systems, so you…
I'm fine with current state. And I know exactly how to make it faster than this initial code -- the possible performance hit is only there when using really large radii; it does not affect existing behavior.
using a superellipse was just an easy way to soften a slight discontinuity I was perceiving at the transition point.
Yeah, superellipse "feels nicer", I was just wondering why exactly 2.1…
I think I've figured out that the math you use is "superellipse of exponent 2.1", right? But why superellipse, and why exactly the 2.1 exponent?
While a rounded corner shape that is close to a…
Very nice! I've added some questions wrt exact math used for the rounded corners (and why the math might be expensive). If nothing else, at least the "what is the formula used and why" should be explained in a comment, I think.
I wonder whether all this math could be simplified, given that color.xyz stays constant throughout the whole function, only the alpha changes based on the rounding?
Curious about these two pow() per pixel (which I guess is the major performance cost). What is the math idea behind this? I think regular "just a quarter of a circle" rounded corners that are done elsewhere in blender would be just a sqrt() per pixel; here you are doing something else, presumably?
Does this also properly hide retiming speed text label indicators when strip height is too small? (if so: yay, good!)