I am still fairly new to the field of rendering and I am looking to optimise the speed of rendering. My scene isnt very complex, but I have to render it in 4K, which still needs some time for an animation. I saw a video , which increases the noise threshold to to maximum of 0.5 or even 1, and then increasing the resolution ( and probably down sizing it again, even though it was not shown).
Is super sampling a widely used method in Blender/rendering? Is there a proper tool to correctly downsize images afterwards, without loosing too much of the detail created?
Or is it more doable/better results to stick to the threshold and find the correct threshold for the wanted time?
Yes, it is quite a common method. As far as I know, it’s even used by some big studios. Though if your final image must be 4k in resolution, you will need some good hardware, as you would need to render in 8k and you could run out of memory easily.
Any specialized image or video editor should be able to do it. Last time I checked, Blender’s sequencer wasn’t the best at this (I couldn’t get good filtering when downscaling), but it’s been a while since I checked.
The extra detail will still help your image be clearer than it would have been even after downsizing.
It depends on the resolution of the image. If your supersampled version has a huge resolution, the denoising time could get quite long and the memory comsumption will greatly increase. Are those disadvantages acceptable in your specific case?