Topic: Lighting Feedback / Request
I have three requests (below) around the ambient light system. This is a great feature, but I think it could use some improvements. I'm always willing to bribe to get my way, so I'll be happy to make another donation if any of this stuff is fixed.
These issues notwithstanding, this is a fabulous project. Thanks to all those who work on it: I have turned several other people on to its magic.
Sarah
1. A real Apply button for the custom lighting settings.
The first is a simple UI request. In the most recent version of the SVP Manager, the custom lighting settings have an Apply button that acts like an Okay button, closing the dialog box. This means that if you want to experiment to find the settings you like, you have to go back through the menus after every adjustment. I would suggest that you follow standard UI design and have the traditional Cancel, Apply and Okay trio. At the very least, please just make it so Apply doesn't close the window!
Making things worse, the menu item is now buried in a sub-sub-menu. The way it overlaps makes it frustratingly difficult to even open it without accidentally closing the submenu. Having to navigate all this every time makes finding the settings you like a real chore.
2. Save my custom lighting settings
Not all types of video really want the same types of settings. It would be great if we could give our settings a name and add them to the list next to the current Type 1–4. (Incidentally, it would be nice to give those more descriptive names.)
3. Higher quality lighting effect
The first rule of ambient lighting is that it should not distract. The lighting area is broken into large rectangular regions. The human visual system is designed to notice high frequency boundaries like this. For horizontally letterboxed video, the effect is so bad as to make the feature more or less unusable.
This can be easily solved by using any reasonable interpolation algorithm rather than the current nearest neighbor approach (nearest neighbor arguably has no place in any kind of image processing ). Preferably the interpolated values should be subject to a true random dither: patterning can be quite apparent for large-area gradients like those used here.
I'm actually quite interested in this feature, so I would be potentially willing to take a shot at improving the algorithm if somebody can point me toward the code for this feature. (I assume this project is open source, but I didn't see any links to the sources. And I don't speak Russian…. )