Hi all - as a first question, I'd be interested to know what approach others are using to replicate digital sensor noise when comping elements into digitally sourced footage.
Also are there any particular characteristics you are noticing and having to reproduce? (have you seen specific distribution patterns in the luma/chroma ranges, is digital noise more forgiving/less noticable than film grain?)
A note of the source camera you're matching would be good.
cheers
Add comment viewed 208 times Latest activity 7 months ago
Hugh Macdonald
[ Editor ]
Matching to the Red camera was interesting...
Looking at the noise showed very fine luma noise, and large soft chroma noise in R and B.
I ended up matching it using Shake's FilmGrain node. Two of them, one for luma and one for chroma. It did a pretty good job.
Usually I'll do the grain using a combination of FilmGrain nodes, straight-forward random noise and samples of noise from flat areas that can be looped and duplicated. In the end, it's a matter of whatever works.
In response to the other answer (suggesting Nuke’s Grain2 node), I’ve just had a play around with it, and I’m not sure I’m seeing the flexibility that was implied. It doesn’t seem to have any control over the channel separation in the way that Shake’s FilmGrain node does…
Boris Lutsyuk
[ Editor ]
Nuke. "Grain-2" node give an expecting result. You can control anything to get best result. And also never forget to control the Blue channel :)
julik
[ Editor ]
I've described a technique for that that works very well for me in this fxguide tip
If you want to be 100% on the safe side it's good to have the DOP/cam crew shoot you a neutral black plate (1-2 seconds with the lens cap closd) with some gain (to make it gray) or a gray card.
Boris Lutsyuk
[ Editor ]
If you want exactly emulate the film grain - you need to control "black", "grey" and white zones - because intensity by this zones is very different. especially if you will look at the each RGB channel. Get the reference material from camera and compare.