Can someone explain the logic of the Render Resolution Percentage in the Properties->Render options? This makes no sense to me and does not work as expected. For one thing, changing the percentage to 50% still outputs an image at full rez, while the Image Viewer displays a full-size image that is cropped 50%. Worse than that, setting the percentage to anything other than 100% messes up the composite in numerous ways. It changes the size and position of roto-mattes, and it changes the motion and position of transformations controlled by tracking curves. Oddly, it does not affect a Planar Track node.

I have read other posts that suggest placing Scale nodes after each image source, and apparently each mask, track source, and transformation node in order to accommodate changes to the render resolution. That is insane to have to place a scale node after every operator in the script.

Why does it not simply reduce the size of the final output like any normal scale operation. Why does it have to affect everything in the composite? There seems to be no end to the bizarre logic at work in this software.

  • 8
    $\begingroup$ do you have a question in there, or is this just a rant? $\endgroup$
    – David
    Jan 30, 2018 at 1:28

2 Answers 2


Render percentage is meant as quick way to get a smaller faster preview of the final image while maintaining proportions of the final size at a fraction of the render time.

Unfortunately the current compositor is not canvas aware, as as such does not work well with differently sized images. It is meant as an extension of the Render pipeline, and as such it expects images of the same size.

If you plan to use the compositor then you must either not use render percentage, or if you do compensate by scaling as you already mentioned.

Can't quote any sources at the moment but I believe I have read somewhere in the developers mailing list, there were vague talks about the remote possibility of eventually improving this situation for upcoming Blender 2.8# series, though nothing is certain at this point.

  • $\begingroup$ Thanks for the clarification. After some experimentation I got it working correctly. $\endgroup$
    – zippy
    Jan 30, 2018 at 3:47

For a comp with only 2 input images and 1 roto mask, I had to add 2 scale nodes, 4 math nodes - 2 for each transformation node x/y, and 1 value node to feed the scale factor into the math nodes. Whenever I change the render resolution I have to modify the scale factor in the Value node to properly affect each transform. I suppose once you know the rules, it can become part of the process of building the comp, but things could get ugly pretty fast in a large comp.


Not the answer you're looking for? Browse other questions tagged .