I’m doing an external optical flow analysis and piping the output of that into Notch with a VideoIn node (Spout currently, but later through a video capture) - this is in the same format as the Optical Flow node within Notch (RG vector image).
However, it appears that the Motion Video input is not really accepting anything except Optical Flow nodes or pre-rendered video. Can anyone verify this doesn’t work for them either?
For anyone interested - the input format of the motion vector ports (image emitter/image field affector) is a signed 32bit float, which is tricky when you are trying to import from an external Video In source (a very large buffer we wanted offloaded from the main renderer) which generally caps out at a 10bit. We ended up compressing the data into a 10bit frame and then unpacking and signing/remapping with a HLSL shader. As it was only vector data the information loss wasn’t too much of an issue.