Vector Blur error.

Hi, I’m trying for a while to create realistic motion blur, but Vector Blur gives the same result as the Ordinary old and slow Motion Blur : A blur effect Before and After the moving object.
In the real world, as the motion blur is due to the visual persistence of the eyes, there is never a blur before the moving object, because the object has never been there at the moment the eyes see it.
So, is there a way to get rid of the blur effect visible before the moving object ? I encountered the problem recently for the blur of an airplane propeller, and now for a firework effect.
Thanks in advance.

But this is exactly what a realistic motionblur would do, having a blur on both sides of the object.
The other thing is some kind of smear effect.
Vectorblur on faster objects on the other hand is very limited, because the objects have still a shape in the center, which is not very realistic.
For an Propeller of an airplane I would choose the Transparent texture version in Blender.


I don’t agree with you. Motion blur in real life gives a blur in the full area of the moving object and in it’s trail. Never before the object in the area where it hasn’t yet been.
In Blender vector blur, the object is in the middle of the blur effect. In a still picture, you often can’t tell what is the direction of the object, because the blur is symetrical !
The real effect is similar to the directional blur, except that the blur follows the moving of the object. Real motion blur is like a smear effect done in photoshop, following the object.
For the propeller blur, I have already used textures, but textures works only for rotating objects, and I’d need a tool working in any case.
I’m not a programmer and I cand do it myself, but I think that the vector blur node could be modified to match the real phenomenon.

Roubal: I don’t mean to be dismissive, but this topic has come up several times over the years.
Here’s a recent example, but earlier threads have gone into more detail. I recommend a search.

It turns out that in “real” motion blur, (as in from a real camera) the blurring does occur on both sides of a moving object. It’s to do with when the shutter opens and closes. They’ve looked into it in quite a lot of detail.

The trouble seems to be (it was for me anyway) that some artists expect the comic book type motion blur that you describe, where the blur is entirely behind the moving object. In comics you use motion blur to indicate direction. In animation you use it to smooth motion.

If you disagree, please feel free to post some real world references of the motion blur you describe.

Vector blur ignores any camera movement, only object movement. Vector blur ignores framerate.
Motion blur accounts for all movement and framerate and shutter speed of camera and objects including rotation and accelleration. It simulates the exposure for the duration of shutter open. You see in 3D view the position at the start of the frame, but during the frame the thing moves, and is captured on the film - hence the blur appears to be in “front” of the object.
Directional blur kinda simulates lateral camera movement, although it ignores depth parallax and rotation.

Whant I really need is a directional blur, aplied to moving objects.
It would be a mix of The motion Blur filter available in VirtualDub,and the directional blur.

Sorry for the double post, I have been cut while trying to edit the previous post, and I haven’t been able to delete it.
What I really need is a directional blur, aplied to moving objects.
It would be a mix of The motion Blur filter available in VirtualDub,and the directional blur.

If you disagree, please feel free to post some real world references of the motion blur you describe.

@Freen : I think that it is an error to compare with cameras, even if in my opinion, real cameras act more like human eyes than CG cameras.
The “tool” we use to look at the world is the eye, before being the camera. So, our vision of the world through a camera is the combination of the camera and the eye’s properties.
This combination gives the motion blur effect. It is not only an effect due to the shutter behaviour, it is mainly an effect due to the persistence of human vision, decreasing with time. It is not only a special effect used by artists drawing comics !
Also, the worlds hasn’t necessary to be shown like seen through a camera. The camera is only a medium,and we should forget it. The main goal in my opinion is to show the world as it is. So if I want to show it in a realistic way, I haven’t to mimic a camera behaviour, but only the eyes behaviour.
The persistence of vision has to be faked, because human eyes perception is a continuous phenomenon. In a CG animation, the CG virtual camera has a framerate (24, 25 or 30 images per seconds) and no persistence. As movement is recorded in a discontinued way, the persistence of vision of the real human vision as to be simulated.
The blur on a film depends of the time of exposure of each frame. If an object moves in the field during the time the shutter is open, it will be blurred more or less, function of the time of exposure. But the “trail” will always be after the object.
If the shutter time is short, the frames will not be blurred on the film. The effect will look like a stroboscopic effect like the Motion blur in VirtualDub : a serie of instances of the objects, with decreasing intensity due to the decreasing of the eye persistence in time. If the speed of the objects is low, the trail will be short (overlapping instances of the object), and if the object is fast, the instances will be distant from each other, giving a trail made of several instances of the object with decreasing intensity, but no real blur effect.
The opposite (trail before the object) is a phenomenon only visible on a still photograph (typically a car with front lights switched on) taken at night with a flash released at the opening of the first shutter curtain. It is a very special case, and can be corrected using a flash released just before the end of the exposure (named second curtain flash).
So, it is not possible to show through a real video the real motion blur seen by the eyes, without adding the blur with an appropriate filter !

This is not true. Blender’s MBLUR button simulates the way it hasppens on film as closely as possible. The only difference is that on film it wil be one continuous streak instead of a series of several steps.

What MBLUR does is, in essence, called frame blending. The shot is broken down into sub-frame time samples which are then divided into equal alpha samples (1 / number of samples) then composited together via alpha over. The heaviest alpha weighting being where the trailing edge of the object catches up with it’s leading edge of it’s leading edge in subframe time. The heaviest alpha being in the area where the object overlaps itself (in the center of it’s travel across that particular frame). Exactly the way it happens on film.

For the effect you are looking for, vector blur would need a toggle button(s) to allow disabling calculation of previous or next frame samples. You can simulate this by doubling the ipo’s length then duplicating every other keyframe then moving the key one frame forward in time. Then on the render tab bump up the frame step field to 2 so that only every other frame gets rendered. Obviously, this is an incredible pain in the ass but for now it’s all we have.

For MBLUR the devs would need to allow us direct access to individual MBLUR samples or at the least the ablilty to assign algorythmical weighting to the samples the way we can with textures (quadratic curve, cardinal curve, or sinusodial curve…Blender currently assigns them in a linear fashion so that’s already on the table) or better yet we would be allowed to draw a custom curve to define the weighting. The samples are already being stored in spare buffers somewhere so I don’t think it would be too difficult to give us access to them. As many times as I’ve seen this topic brought up I’m surprised some Blender pro hasn’t coded it yet due to a need in their personal work. Maybe Shake and Fusion are just easier than coding at this point. Trails, echos, and their associated derivetives are not just common in professional work, they hold a place of prominence (no SmallVille episode would be complete without multiple instances).

Hopefully in the near future Blender will make a quantum leap in her time game capabilities, i.e. nested frame rates for scenes, compositor image inputs and vse strips. Such a feature would allow for effects such as curved vector blur using multiple sample positions which are then frame blended together from 120fps back into a 30fps scene (4 interpolated samples per vector blurred frame in this example…currently limited to 1) meaning that were getting the best of both worlds: the same accuracy of object interpolation that’s currently only associated MBLUR with the beatifully streaked sample rates that are currently only available via Vector Blur. This can still be done through Blender now but the solotions are anything aother than elegant.

You’re right, the ability of disabling calculation of next frames would be a great improvement of Vector blur.
For my own, I have never been able to learn an other language than some QuickBasic basis, almost 20 years ago… so, I can’t do anything to improve it myself.
I hope that this feature will be available in the future.
I have used the method based on video tracks duplication and shifting each track 1 frame, while reducing the alpha value. It is useful in some cases, but doesn’t interpolate between frames, so it is efficient only for slow moving objects… when the image of the object is overlapping. It is not a real smear effect.
Thank you for your explanations.

I’ve been hoping for it for well over a year myself. I have the same no-geekus-speekus problem myself. I have a friend who is a C programmer and he’s willing to teach me but I don’t have the time to set aside on a regular basis; this has always been my stumbling block.