[Feature request] Subframe sampling for SlitScan



In general, Natron’s subframe accuracy works well with generative tools like the Roto or the Shadertoy node, as TimeBlur shows. Could we have SlitScan exploit this feature?
Can’t wait to see what people would create with this.

(Frédéric Devernay) #2

This requires precise retiming. Your best option is to retime your video to get more fps, and then you can slitscan it. It could be done inside slitscan, but would be computationally too heavy.
There are free (eg slomovideo) and non-free options for retiming out there. The non-free ones are still much better.


Indeed. But the precise retiming, ie. sub-frame sampling is what could be done with the generative nodes mentioned above, no? It would be utterly memory-heavy but TimeBlur could sample the upstream with a different time delta for each pixel if need be.

(Frédéric Devernay) #4

the basic operation in openfx is to render a rectangular region of an effect at a given time. If every pixel has to be rendered at a different time, this would take an enormous time to render, for many reasons. The only viable option is to retime the source videos (not necessarily the input to slitscan) at a higher framerate.
My guess is that what you want to slitscan is actually a fixed image. then just increase the fps of your project and slow down the motion, you’ll get the desired effect.
The goal of these tricks is to cache more and compute less.