Blur node

[dgruwier]

A node that blurs a given input along the object UVs.
I'm imagining having a shading network with noises, curvature, the whole procedural shabang, and adding a blur node at the very end to blur the entire thing, perhaps with some filter type options.
It would be functionally the same as baking out the texture, blurring it in a different application, then importing the result, but it would be live in the shading network.

I'm not sure if this is even possible, given that, as far as I know, nobody seems to have implemented in a renderer (?). But if it is, it would be a sweet tool for creating procedural materials.

[seandunderdale]

Seconded. I would love a bunch of decent blur tools in Maya / Vray. No idea how much it would impact render time having to evaluate blur filters in a shader network but the idea sounds great.

[yolov]

After a quick chat with our devs, this might be possible, but there's a good chance it's going to be quite slow.
I'll put a note to consider this.

[dgruwier]

yolov Thanks. I was afraid of that. I was hoping there might be a fast way to do it, but I suppose it would have to be handled similar to other types of raytraced blur.

[BeneZ]

Blur node - great idea - what happened?
No matter which speed, sometimes it needs a slightly final blur for noise or whatever.
Thanks

[yolov]

Noted, I'll add your vote to it in our system.

[Mihail.Djurev]

Unfortunately, there are issues with blurring. Doing blurring on the fly would have a significant performance hit, especially if there are multiple blur nodes in a graph.
Another approach would be to bake the input textures at a fixed resolution. However, this approach has its own drawbacks:
* Fixed resolution
* Textures relying on 3D UVs won't work.
* Triplanar mapping won't work
* V-Ray Dirt won't work
* All textures based on geometry data won't work: UserColor, Random by instance, etc.
* It would require a precompute step

[BeneZ]

Unfortunately, there are issues with blurring. Doing blurring on the fly would have a significant performance hit, especially if there are multiple blur nodes in a graph.

In my case and from my experience even a bad blur can be saving you from long processes with export, import etc. without any chance for preview.
I think a simple blur with a warning ("hey, use just in case of emergency, not recommended will hit your performance") would be better than no blur