Corona Renderer for 3ds Max > [Max] General Discussion

How is Anima able to get motion blur to work with it’s 4D motion scanned people?

(1/1)

UltraMagnus:
How is AXYZ’s Anima able to get motion blur to work with it’s 4D motion scanned people which have changing topology?

Corona isn’t meant to support mb with changing topology but how is Anima overriding this with functionality going down to very low shutter speeds.

I understand that if the shutter speed is high enough then corona will still be able to generate mb as there is enough information.

But, with anima you can take the shutter speed as low as you want and get no error. Taking a closer look, mb looks accurate up until a certain point but when it starts to get low it’s like a hack seems to kick in to replace and imitate normal mb? You can see how the geometry begins to crumple up in the buffer but it’s still very convincing even at 0.5! Pre generated motion vector data? If you take the shutter speed really really low then things start to get crazy.

Another interesting thing is that if I move the timeline slider while interactive rendering the meshes start to deform (see attached image) but this happens even if motion blur is off. I can even grab the actor in this state and convert to edit mesh before stopping interactive where it goes back to normal.

maru:
I am not familiar with these animated models, are you sure that they indeed have "changing topology"? (changing number of vertices/faces etc)
Maybe it's just surface deformation with the same number of vertices per frame? In that case, motion blur is expected to work fine.

And, as far as I remember from some old discussion with the devs, even with changing topology, motion blur may *sometimes* work correctly depending on the specific case.

UltraMagnus:
Hi Maru thanks for the reply, see here https://docs45.axyz-design.com/article/4d-digital-humans/

Scroll to the bottom and you'll see that it mentions changing topology. The uvw map position also changes at intervals as it plays through and they are streaming 100s of texture maps which seem to completely change correlating with when the uvw mapping changes. So I don't understand how it's also blending them together when using such low shutter speeds.

I did some more tests which are interesting. The two people on the left are the 4D people, the 3 on the right are traditionally rigged people. At extremely low shutter speeds strange things start to happen to the 4d people where as the rigged people are ok.

maru:
This probably means that we are able to somehow estimate motion blur based on the provided information, but it may fail in some other cases.
Something similar happens with volume grids (we probably can't talk about changing "topology" here) - a different volume grid file is loaded per frame, but we can get correct motion blur from that.

As far as I understand the point of this thread is that this is not working as expected with Anima people, so we should contact them and try to fix it. :)

UltraMagnus:
Well actually I was just personally quite curious how they were bypassing mb on changing topology because at low shutter speeds it's like a different type of mb (slightly more jaggedy) starts to kick which does a good job of imitating normal mb… until you go super low shutter speed and the models spike out everywhere. Some type of override, black magic going on here hehe, who knows.

I don't usually go to extreme low shutter speeds so I actually only noticed how they exploded everywhere while doing some testing for this post. But yea I guess that is something that needs to be fixed if people are using them with very low shutter speeds.

Maybe I'll ask them how they do their mb, although they only have support tickets, not sure if they have time to explain how their software works without me actually having a problem lol.

Anyway thanks for chipping in and offering your thoughts :)

Navigation

[0] Message Index

Go to full version