Long shutter motionblur, only trail behind, not infront?

[fixeighted]

Hehe, well I started to type a really quite long post about that exact thing - I revised it a number of times and eventually gave up, as it
is confusing even having to try to explain all the confusions between everything.
The bias definitely works but it's no good if it can't be applied in a camera setting.
At the moment, other than standard short motion blur, I'd look for other solutions, as this is just too much of a time-sink
with no guaranteed results to be worthwhile.

[kosso_olli]

The bias definitely works but it's no good if it can't be applied in a camera setting.

Oh wow, now I get what you mean! The view can not be a camera, it has to be a perspective view? It would never occur to me to even try it this way...

At the moment, other than standard short motion blur, I'd look for other solutions, as this is just too much of a time-sink
with no guaranteed results to be worthwhile.

I wish it was that simple. We had to do this quite a lot in the past with head- or rearlights of moving cars. If it could be done in-camera in V-Ray, it would help immensely.

[Buck]

Buck If I understood correctly, you would like for the motion blur to be lighter at the front (where the object is), rather than all the way through? A negative Bias (Overrides>Motion Blur) parameter should be able to help in this case.

The main thing I can't seem to do is get the motionblur to happen only behind the object when using a very long amount of motion blur (100+ frames in my case). I've fiddled around with it a good bit and haven't found any way to do it. Not sure if it's possible or not.

[^Lele^]

The main thing I can't seem to do is get the motionblur to happen only behind the object when using a very long amount of motion blur (100+ frames in my case). I've fiddled around with it a good bit and haven't found any way to do it. Not sure if it's possible or not.

You mean it works as intended for shorter blur lengths?

[Buck]

You mean it works as intended for shorter blur lengths?

to be honest I didn't mess around with it at lengths shorter than 1 frame. Ill run a test today.

[SonyBoy]

Why are there two locations for parameters regaring the motion blur: One in the camera itself, one in the camera render settings. Some values even overwrite each other. Confusion times two.

Ditto this.

[^Lele^]

Also, one thing that always give me headache: Why are there two locations for parameters regaring the motion blur: One in the camera itself, one in the camera render settings. Some values even overwrite each other. Confusion times two.

Both Max and Maya have standard cameras (i.e. non-physical), and at least for Max there is no Modifier that adds the camera attributes (while in Maya it's present).
Having global attributes for moblur helps then to achieve things like motion-blurred Ortho views.

[fixeighted]

What I personally would love to see is one of the minds that created this awesome tool present a way of just doing this, rather than us all spend 3 days and possibly longer
discussing how frustrating it is dealing with the interface
If we are missing something then that's cool - I have always been here to learn best practises, where they exist.

For this particular thing, essentially controllable light trails, there are plugins that do it in various DCCs, XParticles for C4D for example is a modern one, KY Trails I mentioned is an old one that still works.
Still more options in AE, Nuke etc.
Search for this sort of thing and there is practically nothing but these, so clearly there is an issue with 'real' mblur that may be difficult to overcome without using 'fakes'.

However, because we all now want everything to be in one pass, ideally, we all imagine and expect increasingly capable tools to do that.
To my mind, simplistically speaking, it should be a matter of being able to, by coding, dictate where the blur starts and how long it is, as well as how and if it tapers, plus where the brightest spot is.

I'm no coder, so I have no idea how difficult this is to effect....please enlighten me

[kosso_olli]

However, because we all now want everything to be in one pass, ideally, we all imagine and expect increasingly capable tools to do that.
To my mind, simplistically speaking, it should be a matter of being able to, by coding, dictate where the blur starts and how long it is, as well as how and if it tapers, plus where the brightest spot is.

I think the same, but there should also be a different way to get the result: By using plausible real-world values for exposure time, F-Stop etc. For long exposure shots, this does not work. Maybe someone of the Chaos staff has an idea how this can be done.
fixeighted
Do you have the results from Corona at hand?

[fixeighted]

I didn't save any files, as it got too confusing whilst trying to work with it in Vray.
The result is the same walking 'Vlado' image on the first page, where you can see the settings.
As it turned out, it worked for that in a way, though it also does suffer from some of the same issues as Vray in other situations.
What I loved about it is that it literally took seconds to figure out, whereas I just fought with Vray for 2 days to work out what may or may not work.
The process was exhausting

I found this photographic technique which perfectly illustrates the effect I believe, and it's all in-camera.
I have hope therefore that it is translatable to a 3D camera.
https://www.canon-europe.com/pro/sto...s-motion-blur/

[kosso_olli]

Tried to tackle this once again, but failed. It is a mess.
The bias value does not have any effect whatsoever, neither in a free perspective view, nor in a standard or physical camera.

Also, using the same values in a physical camera and in the camera settings yields different motion blur result. A shutter value of 100 frames in the physical cam results in quite a long blur, while using the same value of 100 frames in the Vray camera settings results in a considerably shorter blur. All the other settings were left at default. Could anybody please try to confirm this?

Physical camera with duration of 100 frames:


Standard Max camera or perspective view with duration of 100 frames from the V-Ray Camera Settings window:

[fixeighted]

Ah ok, so you are using the physical camera rather than the vray one. Not sure what the exact differences are but yes, it's still unfathomable.
Regardless, from a UI simplicity standpoint, I would argue that any camera options need to be in the camera rollout, as splitting them makes no sense to me - just adds confusion.
The extra ones can be there but they need to be documented properly, as being additional as a utility where other means don't exist, as Lele intimated.

My simple question is how is there motion blur before an object has even started moving?
All other questions come from that first point I think.

The effect we are looking at achieving, as in that link, is called 'rear curtain sync' - it can be 'front curtain sync' also - and countless examples are easily found, everywhere, because it's an artistic choice and a real-world capability.
This surely must be possible in cg, although I can't find any info on it.

On an only slightly separate note, I wanted to see how camera motion blur affected this example.
I animated the camera travelling from right to left, with the ball stationary.
Camera motion blur doesn't appear to work at all.

This simple scene is just really bizarre...try it and see....it is totally counterintuitive and exhibits the non-working camera mblur as well as the other anomalies.
https://www.dropbox.com/s/d9as8e5lhs...MBLUR.max?dl=0

What this says to me is that software looks at an animated object...sees that it is going to be in position x at a certain time, then calculates the motion blur and just sticks it on the top, regardless of any other condtions.
So at frame 0, when there is no movement at all, it adds all the blur for the specified time or relative shutter speed.

I want to be wrong...I want to be told I know nothing about real world cameras and that it's user error. But the more I struggle with it the less likely that seems

[fixeighted]

As usual, I answered at least one of my questions immediately after posting above.
Blur is in front because it is of course capturing the image over a frame range and exposure duration.

So the effect we want is still the curtain sync, which may or may not be possible.

Corona alters the blur very simply with the shutter offset.
Bias in Vray seems to have some effect but it is almost imperceptible.
Vray shutter, blur duration etc. - unfathomable in comparison.

Again, these settings are extremely hard to gauge without the IPR updating accordingly, so I can only hope that'll be addressed
Using Corona, it is instant (sorry to repeat myself here) and therefore user friendly.
For each and every setting change in Vray I need to stop and restart the IPR. It's too frustrating and at this point I pretty much give up trying.

Here are my tests, with Corona. I wanted to do a Vray comparison but I simply can't replicate it, nor come even that close, which saddens me.
And as mentioned, in any case the camera blur doesn't work at all, so no comparison possible for that.
https://vimeo.com/474663167

I want an AHa! moment with Vray, so help me/us all out here, you clever developers


EDIT: I just noticed that camera motion blur DOES work, but only using GPU. Still the same non-functioning IPR though - actually worse than cpu as once running it does not update frames even with no changes.

[kosso_olli]

The first image in this post is set up like this:

- sphere moving along a path from the upper left to the lower right, starting at Frame 0 and ending at Frame 100
- the physical camera is rendering Frame 0 with a duration of 100



This image is done the same way, but rendered at Frame 50, still with a duration of 100 frames. As you can see, the trail is starting where the sphere is at Frame 50, and than indeed calculated in advance, "into the future" so to speak. This is an odd way of doing it, and it contradicts my understanding of physics.


All other settings for the motion blur were left at default values.

[fixeighted]

Yeah it's too messy and weird for me. As I mention in the other post, if I had a job on including this then I would jump to Corona to do it.
It's not ideal to switch, but it's literally seconds to set up and iterate, rather than these days and days of frustrating tests yielding effectively nothing usable/obvious/logical/tweakable....