Not even sure if it's possible to make that work. The point is to send it to the screen upsampled and let the application downsample it to the output resolution. If everything goes right, it will use some sort of filtering (instead of nearest neighbor) to output the downsampled image. To do that, I believe SV_TARGET is what needs it's dimensions changed. Maybe I'm not quite getting the "output" texture type? :frown: Either way, I've implemented a proof of concept of my idea by upsampling each pixel indiviadually and then merging them in the same pass. The result isn't amazing, but on top of SMAA it's pretty nice. Unfortunately, the performance impact of the interpolation algorithm is currently waaaaay too stronk. Just running the Interpolated Downsampling filter is more of a performance impact than 2x SuperSampling. And the configurable "buffer" type textures? *insert, doe-eyed, "please" smiley here* I need that too to get SMAA T2X working.