Velocity: Clamping or not ?

Hi there,

I was wondering if you should clamp the velocity pass or not ?
Workflow: 32bit, floating for AE and Fusion

Thx

You don’t clamp velocity. This way, you can plug it in and get exact velocity of the object. If you clamp it, then you must know the max velocity in the frame and input that as a multiplier. Velocity changes from shot to shot, its not ideal. With unclamped velocity its always contestant. You also don’t need 32 bit, 16 bit half float is more then enough.

settings

Thx a lot, appreciate the quick response.

and what about the other settings ?

a) clamp velocity: unchecked
b) max velocity value: ?
c) ignor z direction: checked/unchecked ?

max velocity needed only if you clamp

z direction is, velocity is calculated in screen space X - horizontal, Y - vertical position of the objects. Z - would in 3d mean facing camera, however most velocity blurs in 2d only accept 2 inputs, and never produce proper results with 3 inputs Z, its really a big limitation of the 2d blur - if you have extreme motion where objects move to or away from camera it will produce artifacts.

Also, crossing motion blurred objects is tricky, you would have to render them separately and blur/merge them in comp. That is why we prefer 3d motion blur it kinda solves those problems at the cost of render time.

And of course - with 2d motion blur shadows will never blur, as well as objects in refleciton/refraction

If you set Max Velocity to anything besides 1.0 you’ll have to remember your value and multiply that in the motion blur plugin.

I don’t believe you can set the maximum velocity explicitly. It returns that value after the rendered frame is done. Its strictly there for the user to estimate maximum motion of object in frame.

cool.. ok.. I think it’s really better to render the MB in 3D… really.. yes, it cost more render time, but considering the headache and comp work.. then I only have to handle the DOF on top of it. Thx for the hint.. you made my day !

‘Max Velocity In Last Frame’ as the name suggests gives you a hint so that you can pick a scalar that will fit everything into 0.0 → 1.0 if you do decide to clamp, however the Max Velocity is a user scalar by which all velocity values are divided.

So if you set the max to 100, you’ll have to mult your velocity channel by 100 in the comp to restore it which is really really bad juju obviously since even the submitting artist never remembers what velocity channel they picked. Ditto with depth channels. Leave them at 1.0, disable clamping and let (ideally multipart) EXR do its thing.