FORUMS: list search recent posts

Motion Blur ... present but not blured

COW Forums : Adobe After Effects

<< PREVIOUS   •   FAQ   •   VIEW ALL   •   PRINT   •   NEXT >>
Nicolas Guionnet
Motion Blur ... present but not blured
on Mar 29, 2019 at 9:36:07 am

Hi,

I must have done something. Everything was fine 30 minutes ago. Now Motion blur appears as the superposition of two positions (in final quality and adaptive resolution) :



When rendered the problem remains ... but in fast preview, the real blur is back ... 😲

Any idea ?

Thanks


Return to posts index

Nicolas Guionnet
Re: Motion Blur ... present but not blured
on Mar 29, 2019 at 9:42:39 am

This might help :



Return to posts index

Oleg Pirogov
Re: Motion Blur ... present but not blured
on Mar 29, 2019 at 10:34:24 am

Could you please elaborate on how the layers are interconnected through expressions? I mean, what references what?


Return to posts index


Oleg Pirogov
Re: Motion Blur ... present but not blured
on Mar 29, 2019 at 10:40:49 am

Add: I ask, cause this sort of things:

(which looks kinda like what you are experiencing) may be caused by cross-referencing + motion blur.


Return to posts index

Nicolas Guionnet
Re: Motion Blur ... present but not blured
on Mar 29, 2019 at 4:52:54 pm

Hi Oleg,
Thanks for your answer.

Cross-referencing ? i am not sure that I understand what you mean. (circular definition ?)

I move the blue gear with the code below that controls its rotation ... (but before you read it) ... I just changed it to a basic time*1000 ... and motion blur is back.

So it seems that here lies my problem :

t= time ;
t0 = 0 ;
t1 = 14 ;
pasT = thisComp.frameDuration/1 ;
SpeedCoef = 5 ;

// rotation speed or derivative
function speed(tt) { v = -SpeedCoef*(tt-t0)*(tt-t1) ; return v ; }

// speed integration to get rotation
alpha = 0 ;
for (var i = t0; i <= t ; i = i + pasT) {
alpha = alpha + speed(i)*pasT;
}
alpha

In fact it is just an innocent way to control the speed of the rotation.
It is strange that it kills motion blur. ... and another details, it is not smooth ... there is a step from time to time ...


Return to posts index

Steve Bentley
Re: Motion Blur ... present but not blured
on Mar 29, 2019 at 5:02:17 pm

That shouldn't be killing the blur but there's a way to check - change your expressions to keyframes and see if the blur returns.



Return to posts index


Nicolas Guionnet
Re: Motion Blur ... present but not blured
on Mar 29, 2019 at 5:12:35 pm

Thanks Steve,

... I replaced this expression by a simple :
time*time*100
... which gives a very funny accelerating effect ... and the blur is back.

So I guess we can conclude that it is the expression's fault ...


Return to posts index

Richard Garabedain
Re: Motion Blur ... present but not blured
on Mar 29, 2019 at 10:13:52 pm

you could always precomp by moving to a new comp...then adding the force motion blur effect...its a bit slower but should do fine


Return to posts index

Nicolas Guionnet
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 2:02:34 pm

Hi Richard,
Thank you for your help.

In fact I tried to pre-comp and use the effect, alas the result is the same. ... except that now, I know it is a possible solution for a similar problem in the future.

So ... thanks !


Return to posts index


Oleg Pirogov
Re: Motion Blur ... present but not blured
on Mar 29, 2019 at 11:53:43 pm

The essence of the problem:

10*time/thisComp.frameDuration;

Motion blur we need.

10*timeToFrames(time);

Motion blur we deserve (just kidding).

While time/thisComp.frameDuration equals timeToFrames(time) at each frame, they are not equal in general, cause the former is continuous function of time and the latter is a step function which takes only integer values. In fact, whole-frame moments are the only moments when those functions are equal, but since those moments are usually the only moments observable in AE, the functions seem to behave the same way.

On contrary, when Motion Blur is applied it needs to evaluate the expression at inter-frame times and that's where the difference between functions plays out: time/thisComp.frameDuration gives continuous values, while timeToFrames(time) gives just two discreet values, the ones that fall into shutter scope.


Well, you don't use timeToFrames(time), but you use this:
for (var i = t0; i <= t ; i = i + pasT) {
alpha = alpha + speed(i)*pasT;
}
alpha

- this has the same flaw as timeToFrames(time): it's not continuous in time but discreet and motion blur has only two rotation states to take samples from. And on the screenshot we see those states "motionblurred".

So, in case you still want to use that expression, you should switch to continuous function to calculate the rotation angle.


Return to posts index

Nicolas Guionnet
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 1:59:36 pm

Thanks Oleg !
G R E A T !
That 's clear ...

I will have to find another way to generate an interesting rotation ...

But ... do we know precisely what AE needs to compute a motion blur ?
The position at every frame is clearly insufficient ...
The position at every thisComp.frameDuration/10 ? ... thisComp.frameDuration/100 ??

Well, I tried thisComp.frameDuration/10 ! ... and it is much better, there is here a decent motion blur ...



... but it seems 10 times slower, while 10 times more computation should not be needed (compared to the case where the rotation is given explicitly by a continuous expression.)

And the motion is not smooth, due to the approximations in my method of computation of the primitive (I guess) ...

So, I think, I will not use a function of time giving the speed of rotation and then computing the rotation itself. I am starting to think that it was one of my silly ideas ...


Return to posts index

Oleg Pirogov
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 2:52:57 pm

>But ... do we know precisely what AE needs to compute a motion blur ?
>The position at every frame is clearly insufficient ...
>The position at every thisComp.frameDuration/10 ? ... thisComp.frameDuration/100 ??
Well, preferably, position at every point on the timeline. But since you have 16 Samples per Frame in the settings, ~16 discreet values per frame would arguably be enough.

On the other hand, your speed is polynomial function of time:
-SpeedCoef*(tt-t0)*(tt-t1)
- why not just compute (I mean, with pen and paper) its primitive integral and use it instead of numerically integrating speed 16 times per frame?


Return to posts index


Steve Bentley
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 3:08:46 pm

I am wondering if the issue is one of "remembrance". There is no memory of the frame before when using expressions. The motion blur algorithm needs a delta between the frame before and the frame to come, neither of which are being calculated on the current frame. I use expressions all the time and get motion blur correctly so I'm not sure what the difference is here. Did the project get uploaded so we could play/



Return to posts index

Oleg Pirogov
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 3:21:34 pm

>The motion blur algorithm needs a delta between the frame before and the frame to come, neither of which are being calculated on the current frame.
As far as I understand it, motion blur just renders the layer at [Samples Per Frame (or more, up to Adaptive Sample Limit)] points in time taken uniformly(?) from the scope, defined by Shutter Angle and Phase.


Return to posts index

Steve Bentley
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 3:25:16 pm

Right but it would have to have a vector to figure out which way the blur should go. And while expressions will still provide that, this one doesn't seem to or is giving an error.



Return to posts index


Steve Bentley
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 3:26:52 pm

I think i'm a couple of posts behind. Once I posted there were a bunch in front of mine that weren't there before.
Did the OP ever post the project?



Return to posts index

Oleg Pirogov
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 3:55:26 pm

>Right but it would have to have a vector to figure out which way the blur should go

I don't quite see why it needs the vector. In my understanding, Motion Blur just overlaps a number of rendered images of the layer mimicing real-life camera behavior rather than interpolates between frames.

>Did the OP ever post the project?

No not yet. Though you might be interested in recreating something similar by simply applying 10*timeToFrames(time); to, for instance, an X Position of some layer and turning on motion blur. I'm convinced it results in the same thing as in the OP's project.


Return to posts index

Nicolas Guionnet
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 4:49:54 pm

>Did the OP ever post the project?

I don't know what an OP is ... but I suspect that it could the guy who started the topic, alias me.

Sure, I can upload the file but as it uses a load of imported .ai files ... I don't know how to do that ... properly.

> In my understanding, Motion Blur just overlaps a number of rendered images of the layer mimicing real-life camera behavior rather than interpolates between frames.

As I said, I tried to change the computation step to thisComp.frameDuration/10, this way providing 10 times more vector images. And indeed AE used them to make a better motion blur. The "sample per frame" option (composition settings > advanced ) determines the number of images that will be used for the motion blur. If you reduce it to 1 or 2 , you get the same poor motion blur issue we were talking about. So I guess AE "tries" a sequence of time values, close from the current time, to compute the needed images through the expression, and then superposes them. If there is only one image per frame provided, it superposes the image on itself ...


Return to posts index


Nicolas Guionnet
Re: Motion Blur ... present but not blured
on Apr 1, 2019 at 4:27:48 pm

Yes, Oleg, of course ...
This is what I did ... finally. 😉


But, in the future, I would have been tempted to use more sophisticated functions (like a Gaussian) ... that require fare more than a pen to provide their primitive ... if this numerical computation had been efficient ...

Thus, I will use simple functions ...


Return to posts index

Steve Bentley
Re: Motion Blur ... present but not blured
on Apr 2, 2019 at 4:36:12 am

You can just drag the project into the message box (where you type) and the COW will look after the upload for you.



Return to posts index

Nicolas Guionnet
Re: Motion Blur ... present but not blured
on Apr 2, 2019 at 9:22:22 pm

Let's try :

13237_machinebug.aep.zip


Return to posts index

<< PREVIOUS   •   VIEW ALL   •   PRINT   •   NEXT >>
© 2019 CreativeCOW.net All Rights Reserved
[TOP]