I am trying to integrate DLSS Frame Generation using the latest DLSS SDK [0] into our rendering setup.
The Frame Generation Docs PDF [1] on page 21 lists camera parameters that can optionally be passed to the DLSSG Evaluation call. The wording for the camera parameters (‘camera’s current position‘) reads like the camera parameters are expected to match the mandatory clip/camera view matrics for ‘the current frame‘. Is my understanding correct that these optional camera parameters only supplement the same information already contained in the mandatory clip/view matrices?
What I would like to do ideally is to pass a specific camera position and orientation from which the interpolated frame should be simulated by DLSSG. However, the description of the camera parameters in the SDK doc does not sound like I could achieve this? Will DLSSG simply interpolate the frames I request/evaluate according to some internal metric, i.e., I can not really specify or retrieve the camera parameters (position + orientation) for a frame generated by DLSSG?
[1] DLSS/doc/DLSS-FG Programming Guide.pdf at main · NVIDIA/DLSS · GitHub