amazing defocus/dof node by eeshlo now in cvs!

i thought i should point this out because the results are absolutely beautiful and waaay beyond anything you can do with the normal blur node! i’ve been cough cough testing it for a while and i’m glad it’s in cvs now! :smiley:

thank you so much eeshlo, you rock!


ps: whoops, i guess i’m getting blind, it was already posted before… sorry! i just thought eeshlo should get more credit for this cool new tool.

Duped thread! Who is this @ndy clown anyway? Mods – do your job!


this is getting nowhere, locked!
and banned andy.

pssst. what is that awesome model you have there?


does that node respects the “masking” of objects so it doesn’t blurs the whole picture but object-based…puh, i hope you understand me

That’s good news!

Isn’t that the model you showed as WIP at the Blender Conference?



I love blender.

EDIT: Samhar: It certainly looks like it does.

oh my good it looks like a much better dof solution than a simple blur.


Yeah it’s very nice indeed. It’s just a shame you need to rerender to update the DoFDist value, anyone know is a override button+value could be added to the node it’s self?


so highly expceted DOF node and so few participants to the discussion? :slight_smile: Strange.

I am still not understanding how to play with set up. Would it be possible to post some examples of pictures and number inputs (or even “shoooort tutorial” :slight_smile: Would be some advanced user so kind? Thank you very much in advance.

maybe it is too hot to play with :wink:

This feature is cool, but it is still missing something… Its not able to be animated!!! I want to be able to change the focal point in a movie, now there is a DoFDist, cant we keyframe the values somehow? Would be sweet if we could do this :smiley:


MicWit: At least DoF dist seems to be keyframeable. Just tried it.

Woweee blender is really getting good recently.


One thing I’ve wondered about that is can’t there be a way to use say an empty say named ‘dof’ for the distance? That way you just need to keyframe the position of the empty. Ideally there would be a point on the camera line that could only be moved along the local z-axis. A visual cue makes it much easier to get the right focus point and maintain it while animating. The reason I said an empty is because it’s easier to apply constraints to it. Of course if the dof nodes can be scripted then you can use a script link to assign the value of the dofdist as the distance an empty called dof is from the camera. Of course arbitrary IPOs would help here (again). :smiley:

Sweet, now I know what I was doing wrong, all good.

The idea from osxrules is a good one, but for now I am happy just to be able to change the focal point throughout the animation even if it takes a few test renders.


This would absolutely rock your socks if some wizard writes some smart glsl that would enable you to watch and tweak defocusing --in the viewport-- , realtime (when in camera view and with on/off button).

Thanks for the compliments @ndy! :slight_smile:

Maybe I’m missing something, but isn’t that exactly how you can already set the focal point? (enable the camera ‘limits’ button to see the focal point)
I simply reused a parameter that was already there for yafray, so it works exactly the same, which also means you can animate it with an ipo.
btw, ‘broken’ actually used the node for his demoreel which you can see on his site, which has some parts where the focal point is controlled by ipo.

Anwyay, yes, using an object/empty was yet another one of the idea’s I played with, but apparently, that I access the scene camera data in a composit node seems to have been yet another mistake. So probably accessing scene objects is not recommended either…
Ton will probably change some things so it might actually become usable.

Absolutely fantastic work. Can’t wait to give this a proper spin.

Can you move it in the scene or can you just edit the numbers and see the result? That is quite hard to control. Having an interactive object is far easier to position.

Yeah it makes sense that the compositor would really have limited access to the scene in the same way that imager shaders do but that info still needs to be passed on somehow and having a two-way relationship between the camera and node makes the most sense.

Are composite nodes scriptable?

I was either reading something Ton wrote to a mailing list or said on irc (can’t remember which right now) in which he said that compositing nodes shouldn’t be aware of scene data, as there may be multiple scenes accessed in the compositor at once. He’s talking about piping scene and camera info to the different nodes, and allowing that info to be saved inside the OpenEXR files, so you can do camera/scene reconstruction effects from just the saved image files, as opposed to needing access to the original 3D files.