Well with that video you basically proved him to be correct, that's nonsense workflow. What he is gonna do, leave it at each spot for few minutes, go for coffee, return, move cursor to other spot and repeat ?
i imagined something like live mask painting directly in VFB
Painting an actual mask solves that. You would paint your cursor at each of those spots and then leave and let it render. It would sample all those positions, not just actual one where the cursor is. He can't use region for it since that would require multiple lasso regions with falloff.
I made a promise to myself to never react to anything you post, and I've successfully lasted for about a year now, but now I will have to break it. Hopefully, I will be able to keep it up for another year afterwards.
A, the entire image is not nearly converged, it has not even
one pass finished. What would be point of using region or brush to selectively converge the image if you would end up covering the entire frame? Practical scenario would be that most of the image is perfectly clean, and some of the highlights are 80% converged, but still a bit noisy around edges, that would not take nearly as much time to clean up as raw render that just started.
B, Okay, let's say we would have ability to paint those areas. Just the sole idea of painting something, and adjusting it during renders require at the bare minimum this amount of new UI elements:
1, Enable brush button
2, Brush size
3, Eraser to remove accidentally painted area
4, Something to toggle view of painted mask
5, Something to erase all the painted stuff with one click
So yeah, let's throw another 5 buttons inside of VFB, entire painting set.
C, While I mentioned use for selective convergence, that's not the primary use of this. Primary use of this is a boost of feedback, You need to see how, say, a lamp looks in your interior, so you quickly hover over it to clean noise there first. It's faster and less clumsy than dragging region around.
Adaptivity should be responsible for noise being even across entire image, so if there are noise hotspots, you should want adaptivity improvements, not complex toolset to "work around" the problem. If Corona did not have any MSI, and most of the renders would contain at the very least few random super bright fireflies on your image, would you be requesting something like MSI, or a complete retouching brush toolself for CoronaVFB? Think about animation. Will you be sitting there waiting for each frame to pop up so you can paint a mask?
