NR at beginning or end of node tree?

I've seen varying philosophies on this, but should NR go first or last? Does the NR algorithm, Temporal and Spacial see noise any better after a grade or before a grade when flat footage is introduced?

Currently I put NR at the end of the node tree before sharpening and range it between 10-60 for Temporal and then half that for Spacial.

Also any tips for how to use temporal vs spacial? What I have heard is the best is to use Spatial at half of what temporal is set to, though I don't remember from what source.


Thanks,
Alex Crocker

A lot of people do it at the beginning so it will cache early. Some people (correctly, I think) point out that if you really squash and stretch the gain, you can increase the noise just in an overall setting, so the NR should come at the end.

I go in the middle, maybe 3-4 nodes in so that an overall gain structure is already established and the noise is already there. After that, I worry about secondaries, keys, windows, clips, and so on, and there's no slowdown because the NR is cached. And I have at least one or two remaining nodes for Trims so we can still do minor adjustments for color if needed.

I go in the middle

That's a good idea Marc! I'm stealing it.

Usually my first node is noise reduction for the cache workflow reasons.

In my experience the visual difference between NR on first node vs last node isn't very discernible. Definitely not worth the re-cache tax for every little tweak.

Qualifier keys are cleaner when NR is early in the node tree.

I’m partial to earlier. It can help with qualifying later. No set amounts. It’s source dependent.


Sent from my iPhone using Tapatalk

A lot of people do it at the beginning so it will cache early. Some people (correctly, I think) point out that if you really squash and stretch the gain, you can increase the noise just in an overall setting, so the NR should come at the end.

I go in the middle, maybe 3-4 nodes in so that an overall gain structure is already established and the noise is already there. After that, I worry about secondaries, keys, windows, clips, and so on, and there's no slowdown because the NR is cached. And I have at least one or two remaining nodes for Trims so we can still do minor adjustments for color if needed.

So right after the main color grade? I usually do a main color grade in the first node, NR in the second, and Sharpening in the third. I don't tend to do all the fancy windows or qualifiers since I can get what I need out of the LGG wheels so far.

I've been adding NR as the first node lately. Though I do work in ACES so that NR is still coming after the Log to (Color Space) conversion.

I don't push things around very much though. My color work is relatively simple. So my thinking is to play with a cleaner image.

i rarely use NR, when i do it goes near the end of the tree, then i know what i HAVE to mash into mush, and use a key so the rest of the image is not trashed

i would never ever add sharpening after turning a image into mush, i'd avoid the need for that in the first place by not useing NR

i have enough horsepower to run any reasonable amount of NR in real time already so cacheing is not really a thought i need to explore in my workflow, and therefore not a limitation imposed, with that removed putting the node at the end is the cleanest option

i rarely use NR, when i do it goes near the end of the tree, then i know what i HAVE to mash into mush, and use a key so the rest of the image is not trashed

i would never ever add sharpening after turning a image into mush, i'd avoid the need for that in the first place by not useing NR

i have enough horsepower to run any reasonable amount of NR in real time already so cacheing is not really a thought i need to explore in my workflow, and therefore not a limitation imposed, with that removed putting the node at the end is the cleanest option

Makes sense. Sometimes I have seen NR go too strong, especially when the Enhanced version of spatial NR is selected (better is better than enhanced sometimes, really?) I have seen this crazy "enhanced" blurring firsthand when editing 12800 ISO footage.