Please do an extended version where you are focusing on something near the camera. Like the back bumper. Yes, it fixes the edge betwen the back and the front but it does not fix the near edge problems !
If you have an appropriate Deep image and a PxF_DeepDefocus with appropriate slice adjustments, you should be able to focus on a foreground object without edge problems.
To quickly grab the focal length/filmback, there is a 'pick camera' button in the Lens Simulation tab. (If the focal length is animated you can create an expression to link to your camera's 'focal' knob)
Pxf_Zdefocus. Would love a camera input pipe pls. What do you think? FYI: zdefocus was originally dev’d at DNEG. I helped beta it many moons ago. Would love to hear your take on 1) how you order motion blur and defocus filters? 3) filtering opinions from CG? I.e in some renderers it can combine filtered and unfiltered(for rendered of cg), and give ok results without dilate/blur hacks, on jumps of distance. 3) the workflow of rendering separate cg layers with filtered and unfiltered depth and them combining them to create one defocus pass. Thanks for your work dude, amazing development & contribution here!
You video has inspired me to change a mergeCG node I have previously written (that combined un/filtered depth), to allow one defocus node downstream, and experiment creating a MergeToDeep node instead. I.e. The main spline is in deep, made from non-deep Apipe images with depth data.
Mixing 'big' defocus with 'big' motion blur is very tricky to do 'right'. In camera defocus is first (lens) and motion blur second (sensor). The problem is that it is hard to elegantly defocus motion vector passes (or motion blur depth passes). At some point you might just want to bite the bullet and do DOF and moblur in 3D.
Regarding the camera input, there is a 'pick camera' button instead. Is that not doing what you want? See 4:37: ua-cam.com/video/hmXCtXeaNCw/v-deo.htmlsi=HaCbGTU3tTkTvMli.
@@pixelfudger >Regarding the camera input In a template situation, centralising a procedurally updated camera asset, allows multiple defocus nodes being adjusted/versioned from a central point in the script (connected with an expression or preferably an asset only dot_connect system). Arguably only beneficial on: feature animation/animatics/large CG shows/multi-shot workflows... especially when the director dials in narrative defocus decisions at the post-via stage. 1) I additionally I do like two extra f-stop offset knobs for Front/Back , for 'creative' adjustment of individual nodes, and then 2) I add an extra tab for sub-frame bokeh vignetting (aka cats-eye/magic defocus) My old defocus node used hacked maths of the OZD node using direct Z.. but your node is significantly better in several areas.. good work dude!
@@pixelfudger > Mixing 'big' defocus with 'big' motion blur is very tricky to do 'right'. yep.. I too haven't found a satisfactory solution. My general rule to reduce artefacts is whatever has a smaller spread of pixels is the top node.
The last time I tried OZD (a while ago) it struggled with foreground objects becoming out of focus because internally it was using 'direct' math on the ZDefocus node. I dont know if it's still the case today.
A Python exception occurred: AttributeError: 'NoneType' object has no attribute 'addAction' File "C:\Program Files\Nuke14.0v5\pythonextensions\site-packages\shiboken2\files.dir\shibokensupport\__feature__.py", line 142, in _import return original_import(name, *args, **kwargs) File "C:\Program Files\Nuke14.0v5\pythonextensions\site-packages\hiero\ui uke_bridge\initNuke.py", line 17, in cacheMenu.addAction(localisation) this happend to me
OK I'm sold. Also the maths in PxFZDefocus is much better than the OpticalZDefocus node
Those PxF defocus nodes with catseye/barndoors/breathing/longitudinal and lateral chromatic aberrations would be superb
Please do an extended version where you are focusing on something near the camera. Like the back bumper. Yes, it fixes the edge betwen the back and the front but it does not fix the near edge problems !
If you have an appropriate Deep image and a PxF_DeepDefocus with appropriate slice adjustments, you should be able to focus on a foreground object without edge problems.
fantastic tutorial
Amazing tools you have going on here man! Thank you for making these, Keep up the good work!! 👏🙌🤝
Thank you so much. lifesaver Gizmo.
Thank you very much for this!!! Would it be possible to add a camera input on the PXF_DeepDefocus?
To quickly grab the focal length/filmback, there is a 'pick camera' button in the Lens Simulation tab.
(If the focal length is animated you can create an expression to link to your camera's 'focal' knob)
Awesome!!! Thank you!
Pxf_Zdefocus. Would love a camera input pipe pls. What do you think? FYI: zdefocus was originally dev’d at DNEG. I helped beta it many moons ago. Would love to hear your take on
1) how you order motion blur and defocus filters?
3) filtering opinions from CG? I.e in some renderers it can combine filtered and unfiltered(for rendered of cg), and give ok results without dilate/blur hacks, on jumps of distance.
3) the workflow of rendering separate cg layers with filtered and unfiltered depth and them combining them to create one defocus pass.
Thanks for your work dude, amazing development & contribution here!
You video has inspired me to change a mergeCG node I have previously written (that combined un/filtered depth), to allow one defocus node downstream, and experiment creating a MergeToDeep node instead. I.e. The main spline is in deep, made from non-deep Apipe images with depth data.
Mixing 'big' defocus with 'big' motion blur is very tricky to do 'right'.
In camera defocus is first (lens) and motion blur second (sensor).
The problem is that it is hard to elegantly defocus motion vector passes (or motion blur depth passes).
At some point you might just want to bite the bullet and do DOF and moblur in 3D.
Regarding the camera input, there is a 'pick camera' button instead. Is that not doing what you want?
See 4:37:
ua-cam.com/video/hmXCtXeaNCw/v-deo.htmlsi=HaCbGTU3tTkTvMli.
@@pixelfudger >Regarding the camera input
In a template situation, centralising a procedurally updated camera asset, allows multiple defocus nodes being adjusted/versioned from a central point in the script (connected with an expression or preferably an asset only dot_connect system). Arguably only beneficial on: feature animation/animatics/large CG shows/multi-shot workflows... especially when the director dials in narrative defocus decisions at the post-via stage.
1) I additionally I do like two extra f-stop offset knobs for Front/Back , for 'creative' adjustment of individual nodes, and then 2) I add an extra tab for sub-frame bokeh vignetting (aka cats-eye/magic defocus)
My old defocus node used hacked maths of the OZD node using direct Z.. but your node is significantly better in several areas.. good work dude!
@@pixelfudger > Mixing 'big' defocus with 'big' motion blur is very tricky to do 'right'.
yep.. I too haven't found a satisfactory solution. My general rule to reduce artefacts is whatever has a smaller spread of pixels is the top node.
Great stuff man !
Thank you ✌🏻
Awesome! I wonder how pxf_zdefocus compares to optical_zdefocus
See pixelfudgers previous video, that also includes pgBokeh?
The last time I tried OZD (a while ago) it struggled with foreground objects becoming out of focus because internally it was using 'direct' math on the ZDefocus node. I dont know if it's still the case today.
Yes it still has this issue. Out of the box your tool gave a better result on the same CG@@pixelfudger
PxF won!
✨👌💪😎😉😎👍✨
by this point, can't you just teach foundry how to design their products? lol
I'm available if they want to hire me. I'd love to do a bug / quality of life blitz on Nuke. :-)
A Python exception occurred:
AttributeError: 'NoneType' object has no attribute 'addAction'
File "C:\Program Files\Nuke14.0v5\pythonextensions\site-packages\shiboken2\files.dir\shibokensupport\__feature__.py", line 142, in _import
return original_import(name, *args, **kwargs)
File "C:\Program Files\Nuke14.0v5\pythonextensions\site-packages\hiero\ui
uke_bridge\initNuke.py", line 17, in
cacheMenu.addAction(localisation)
this happend to me
What were you doing? The scripts triggering the error don't appear to be PxF.