Hey, another good tut as always. I tried convolve vs defocus on a real image. Felt convolve lack control for the amount of defocus we need - like defocus slider in Defocus. The only solution found was to play with the reformat node and adjust the amount. Is there any reason it is not there?
That's true. I would say that the convolve is a simpler node that does one thing, but like you said, you can play with the size of the reformat. To be more practical, you can create a new knob in the convolve, link it to the reformat scale, and save that as a Toolset for future use, kinda like that : set cut_paste_input [stack 0] version 11.3 v4 push $cut_paste_input Reformat { type scale scale {{parent.Convolve1.filterSizeMultiplier}} name Reformat_for_convolve selected true xpos -1035 ypos -266 } push 0 Convolve2 { inputs 2 use_input_channels true name Convolve1 selected true xpos -1035 ypos -230 addUserKnob {20 User} addUserKnob {7 filterSizeMultiplier R 0 2} filterSizeMultiplier 0.2 }
When I was first getting started I worked with a compositor that said he used convolve at the end of his comp to soften overall a crisp high res CG render because he said its much nicer than a blur or defocus to achieve a film-like look, but in hindsight they are actually basically the same at the end of the day right? Any idea why he would think this? Do you have a recommend filter input that'd be particularly accurate to a camera lens sensor?
Hey Matty, the look of a Defocus Node is very similar to a Convolve with a flat disc input, but in the Convolve, you can change the filter to anything you want, not just a flat disc. For example you can use bokeh that you find online, or all sorts of pattern. So to that regard, a Convolve with a "photo realistic bokeh" would look more real than a simple Defocus. Now if the value of the Defocus / Convolve / Blur is very small (I imagine it would be if applied on a whole image), then I doubt you would see the difference, not until you reach a value big enough to see the pattern (Maybe around a value of 3 or 4, under that I don't think I could see the difference myself...) For "simulating" a camera lens, it's a never ending quest, I guess. It starts with the render engine, especially these days, you can do really nice looking defocus with custom bokehs in Redshift, for example. Then in comp, there is a whole list of things you can research on: Chromatic Aberration, Aperture, the different types of lenses, bokeh, cat's eyes effect, lens coating, distortion, then grain, flickering, lens flares, light leaks, etc... Get your inspiration by studying a lot of existing films, pause on frames, and you will find tons of interesting stuff to try to mimic. For the filter to input in the convolve, I usually use a bokeh I extract from photograph either online or that I photographed myself (a tiny LED in a dark room, an old shitty lens pulled out of focus, and you can get awesome results). But when in need of something quick I google "single bokeh" and find something I like every time. (Be careful of the size of your filter input, and resize it very small before processing, or that's the easiest way to get Nuke stuck and crash.) I hope this helps! Cheers.
Thanks for answering the question! I found a good article that goes on about how they used this technique for the lens flares on the Lego batman film from fxguide www.fxguide.com/featured/the-batman-movie-keeping-it-real/
Really informative and thorough. Thanks for this 👍
Hey, another good tut as always.
I tried convolve vs defocus on a real image.
Felt convolve lack control for the amount of defocus we need - like defocus slider in Defocus.
The only solution found was to play with the reformat node and adjust the amount.
Is there any reason it is not there?
That's true. I would say that the convolve is a simpler node that does one thing, but like you said, you can play with the size of the reformat.
To be more practical, you can create a new knob in the convolve, link it to the reformat scale, and save that as a Toolset for future use, kinda like that :
set cut_paste_input [stack 0]
version 11.3 v4
push $cut_paste_input
Reformat {
type scale
scale {{parent.Convolve1.filterSizeMultiplier}}
name Reformat_for_convolve
selected true
xpos -1035
ypos -266
}
push 0
Convolve2 {
inputs 2
use_input_channels true
name Convolve1
selected true
xpos -1035
ypos -230
addUserKnob {20 User}
addUserKnob {7 filterSizeMultiplier R 0 2}
filterSizeMultiplier 0.2
}
@@TheFrameLabGreat! Cheers 🍻
Can you send the that image path you can use in this video
When I was first getting started I worked with a compositor that said he used convolve at the end of his comp to soften overall a crisp high res CG render because he said its much nicer than a blur or defocus to achieve a film-like look, but in hindsight they are actually basically the same at the end of the day right? Any idea why he would think this? Do you have a recommend filter input that'd be particularly accurate to a camera lens sensor?
Hey Matty, the look of a Defocus Node is very similar to a Convolve with a flat disc input, but in the Convolve, you can change the filter to anything you want, not just a flat disc. For example you can use bokeh that you find online, or all sorts of pattern. So to that regard, a Convolve with a "photo realistic bokeh" would look more real than a simple Defocus.
Now if the value of the Defocus / Convolve / Blur is very small (I imagine it would be if applied on a whole image), then I doubt you would see the difference, not until you reach a value big enough to see the pattern (Maybe around a value of 3 or 4, under that I don't think I could see the difference myself...)
For "simulating" a camera lens, it's a never ending quest, I guess. It starts with the render engine, especially these days, you can do really nice looking defocus with custom bokehs in Redshift, for example. Then in comp, there is a whole list of things you can research on: Chromatic Aberration, Aperture, the different types of lenses, bokeh, cat's eyes effect, lens coating, distortion, then grain, flickering, lens flares, light leaks, etc... Get your inspiration by studying a lot of existing films, pause on frames, and you will find tons of interesting stuff to try to mimic.
For the filter to input in the convolve, I usually use a bokeh I extract from photograph either online or that I photographed myself (a tiny LED in a dark room, an old shitty lens pulled out of focus, and you can get awesome results). But when in need of something quick I google "single bokeh" and find something I like every time. (Be careful of the size of your filter input, and resize it very small before processing, or that's the easiest way to get Nuke stuck and crash.)
I hope this helps!
Cheers.
@@TheFrameLab Thanks for that thorough response!!
nice
Thanks for answering the question! I found a good article that goes on about how they used this technique for the lens flares on the Lego batman film from fxguide www.fxguide.com/featured/the-batman-movie-keeping-it-real/
Please