Is there anyone who recognizes this issue and can tell me what’s going on/how to fix it? Its happened in almost every single dynamic vehicle shot I’ve done with Vray. I’m getting very annoyed of how random the quality seems to be.
In this test render you can see the tread of the tire almost like it was completely still when the shutter started rolling then it started to move mid shot, this is not the case in 3d space though as its in constant motion driving down the road.
All of the origins for the objects are in the correct places and location has been applied (usually makes no difference when it isn’t) the motion sampling is set to 8 on all wheels and tires and my shutter speed is set to 60 with the car moving around 100kmh.
Again its very random and from time to time I’ve gotten it to work perfectly, not this time no matter what I’ve tried,
1_2 mile nilufr.blend here’s the scene, don’t unhide any planes unless you want your system to chug a bit.
We don’t work in blender, but the logic should be the same. You should calculate how many times the tire rotates per frame, then use that logic to apply the subframe samples. Additionally ensure that the tire has infinity curve so at first/last frame it continues to rotate into the frames outside of your frame range. Sampling at subframe the objects which rotate really fast requires a lot more than 8 samples. We typically set ours to 32 or 64 in some cases.
Thank you, its been difficult to find examples of this and in the past raising it too high in blender led to insanely high ram usage, i bumped it up to 16 samples per wheel and tire and it looks much much better. It’s just confusing on how i could get perfectly clean results from 3-4 samples and 80km previously.
What’s the wheel’s rotation? We do this for airplane/propeller shots a lot. If for example for a given motor we have a certain rpm at cruising altitude, then we can convert the rpm to rotation in degrees per second, or per frame. Depending on how you are keying things. For us we place key at 1001-1024 so, I’ve asked chatgpt to just calculate the rotation per second (24 fps) and give me the Rotate Z value (our propeller axis).
Here is a quick snippet of that math. Essentially with 0.5 shutter (180 degree) you would need 12-18 samples, so 16 samples is perfect.
With that said, you have to be respectful of the memory here. What we have to do is we also deal with heavy geometry. We actually split the geometry so we only place the rotating parts of the rotor into a separate vray object properties set, and give more motionblur samples to that set, while retaining the default 2 setting for the rest of the geo.
