Better workflow for velocity passes

[SonyBoy]

Setting up a velocity AOV seems to be a confusing and vague process. Can we either get a more automated workflow, or better docs for this? Documentation says we should "locate the frame in the sequence that has the most motion". I for one have no idea what this means -- the fastest motion? Is that in camera space or 3D space? The most overall motion? And even if a clear explanation were available, obviously this is often pretty much a wild guess anyway. Busy production artists don't really have time to test render endless frames to get the reading of "Max velocity in last frame" to figure out the highest number. Could you develop a tool that can analyze a sequence and give the highest reading? Or how about at least a means to retrieve this number after a sequence has been rendered, since apparently the renderer can output this on a per frame basis? This way at least we'd know what to punch in here after rendering a shot once (and since almost all shots are rendered more than once this would be a way to get the correct value. The clamping feature is also extremely confusing. Why do I want to turn this on, or not?

[Buck]

are you having problems getting it to work? what program are you comping in? For Nuke, I just leave the velocity options in V-Ray to the default.

[sirio76]

are you having problems getting it to work? what program are you comping in? For Nuke, I just leave the velocity options in V-Ray to the default.

Can you please post your node setup in Nuke? I've tried in the past but it it wasn't working correctly

[SonyBoy]

are you having problems getting it to work? what program are you comping in? For Nuke, I just leave the velocity options in V-Ray to the default.

My issue is not getting it to work, it's that it seems to be overly confusing to set it up *correctly*. Personally I still often comp in After Effects (yeah I know), and use Reelsmart, but I work with people using Nuke. It frequently seems to be a case of tweaking settings on whatever effect you use to generate the moblur to get it to *look* right, rather than knowing that you're getting the correct result. A lot of this stems from the confusing docs for the velocity AOV. E.g. you yourself say you leave the defaults, but the docs outline needing to extract the max velocity from the frame with the most motion as being the correct workflow. Or, perhaps there is no "correct" way to do it, in which case it would be useful to know that so I can stop worrying about it But TBH it would be pretty helpful to have some explicit docs about how to work correctly with this AOV and common compositing programs including AE and Nuke, bit depth and color management considerations etc. This all seems to be left to the user to figure out for themselves.

[Morbid Angel]

Hey SonyBoy, I made a quick tutorial for you. Note you have to render to multichannel exr, which has to be at least 16 bit half float.