Tweet
Based on what you are seeing, it could be that the EXR export are not scaled/offset as we (RE:Vision) expect them (+ I presume they also won't have per object alpha coverage via that method).
IF Max Displace/Max Velocity (the vector scaling) is actually 1.0 in such case (is the raw untransformed data), then adding 0.5 0.5 to red and green channels would work for us.
But this will only work with a full 32b float pipeline otherwise you will only have motion between 0.0 and 1.0 pixels not clamped.
ELSE If not, then it's a bit more complicated and I can review per app the best solution.
http://revisionfx.com/support/faqs/m...n_vector_math/
I also note as it's frequent question here that in Nuke be careful if you do that (add 0.5,0.5) and use a constant color node to change the frame range in that node to sequence range otherwise it will motion blur only frame 1. (add constant color then set coverage matte)
While I am here, you can just set the Max Displace (it's actually just a scale if you are in float) to image width if you export 16 bit integers (short) png or something (that takes a lot less storage place) until at least 2K that's way enough precision. Then you are not worried about remembering that number.
Pierre at revisionfx.com
IF Max Displace/Max Velocity (the vector scaling) is actually 1.0 in such case (is the raw untransformed data), then adding 0.5 0.5 to red and green channels would work for us.
But this will only work with a full 32b float pipeline otherwise you will only have motion between 0.0 and 1.0 pixels not clamped.
ELSE If not, then it's a bit more complicated and I can review per app the best solution.
http://revisionfx.com/support/faqs/m...n_vector_math/
I also note as it's frequent question here that in Nuke be careful if you do that (add 0.5,0.5) and use a constant color node to change the frame range in that node to sequence range otherwise it will motion blur only frame 1. (add constant color then set coverage matte)
While I am here, you can just set the Max Displace (it's actually just a scale if you are in float) to image width if you export 16 bit integers (short) png or something (that takes a lot less storage place) until at least 2K that's way enough precision. Then you are not worried about remembering that number.
Pierre at revisionfx.com
Comment