Is there a way to get render output size in the node editor?

Hi, is there a node or some way to get the dimensions of the render as an output?

I want to be able to scale an effect based on on how big the render is.

Eg a 10 pixel blur is good for when I’m test rendering at 400 by 300 pixels, but when I’m doing a full render at 800 by 600, I want the blur to be 20 pixels.

Is there a way to achieve that?


Is there a way to write your own compositor node in python or something like that?

Unlinking certain effects from fixed values is a code project for Blender Foundation. The expectation is that these kinds of effects come % of dimensions instead. Otherwise I can’t think of a way to derive the dimensions sorry.

You can use resolution through drivers. Right click on the field you want to depend on resolution (blur size for example) and choose Add Driver > Manually create later. Then from Curve Editor window, change window subtype to Drivers, find the driver from lefthand tree view and click on it. I the righthand properties panel (N shorcut key), choose the Drivers tab and in there set it up as follows:

  • Type: Scripted Expression
  • Expr: the expression you want, for example res_x*0.05 (res_x is the variable name here, see below)
  • under variables list change the variable type to Single Property and give it a meaningful name, res_x for example :slight_smile:
  • in variable Prop field, change type to Scene and select your active scene in combobox
  • copy the resolution path (X or Y dimension, whichever you need) by right-clicking in Render panel on the render resolution value and choose Copy Data Path
  • paste the previously copied data path to the Path field in variable panel
  • if everything is done correctly, your render resolution now drives your value.

For drivers to work you must enable python scripts autorun in user preferences. Open the preferences and in File tab, check the Auto Run Python Scripts checkbox.

This is a logic I personally strongly dislike. Fusion uses this, naming it resolution independence, but what it actually is is total brainf*ck. It is much more flexible to give precise values in pixel dimensions and expose different variables through expressions, drivers or whatever. If I need to depend on resolution I can hook it up in an expression, but most of the time I don’t need this and it will be very counterintuitive to use something like 0.003% in transform or blur instead of 10.5 px. Pixel size value is an exact, quantifiable unit, resolution dependent percent is not. Just as saying that something is a 300dpi image says nothing about its actual size or quality.

1 Like

To illustrate the previous post, most of the relevant stuff is shown here:

1 Like

for example Blur filter, once you check “Relative” turns pixel values into % values :wink:

@kesonmis I agree pixel values are idea for easy repeatability. My preference is a switchable mode, so that you can enter a percentage then switch to pixels and see a more specific value.

You guys are awesome. Gonna try this when I’ve got a minute.
Thanks a lot.