Hello,
is there already any solution for this problem.
Our problem is that we have to render a high quality model to a SceneCaptureComponent2D, due we need to display it within a 2D widget.
Caused by the lack of Temporal AA the model has not the desired quality.
Hence SceneCaptureComponent2D is not an option as long as there is no suitable post processing.
So we started to evaluate some other solutions.
1.) The first idea is to find a way to render the model after the 2D Widget. So the 3D scene just contains the model.
The viewport is first filled with the 2D Widget and afterwards it is blend with the model.
This sounds like a fragment shader manipulation.
Has anybody an idea to get an approach for this?
2.) The second idea is to modify the C++ code and recompile the engine.
This is more complex and I do not know whether the Epic team is currently working on it.
So far i found the corresponding classes within SceneCaptureComponent.cpp, TextureRenderTarget2D.h and TextureRenderTarget2D.cpp.
Additionally the class FRCPassPostProcessSSRTemporalAA in PostProcessTemporalAA.h/cpp is of interest.
Can anybody give me a hint where the code starts the post processing for an UTextureRenderTarget2D, so that I can simply add a **PostProcessTemporalAA **instance?
Thank you in advance.
Edit:
Due Temporal AA needs some frames (default is imho 3) to achieve a time based anti aliasing, this seems to be the reason why SceneCaptureComponent2D does not support it.
Hence when we want to modify the C++ code, the SceneCaptureComponent2D has to store the last 3 frames.