View Full Version : Major big bug...

06-28-2005, 03:22 PM
...as far as I'm concerned. If you have an object enveloped to dissolve over 1 frame, motion blur on, and antialiasing set to 2 passes or more, the object will render slightly (probably 50%) transparent on that frame. It does this even if the camera and object are not moving. Do I win a LW T-shirt? Let me know if you guys get the same results. Thanks.

06-28-2005, 03:37 PM
Seems correct to me. Motion blur is going to try to "blend" where the object was and where it is, and dissolve is included.

If you don't like it, go to General Options (shortcut "o"), enable Fractional Frames, and put your 100% dissolve at frame 30.1 (or whatever base frame it is), which will make frame 31 completely dissolved. Be sure to set Tension on the spline to 1.0 on each key, of course.


06-28-2005, 04:25 PM
No bug, when LW renders a particular pass, it doesn't care if it's moving or not (unless you're using 'vector' blur) it only cares where it is, and renders it there no matter what it looks like...

06-28-2005, 10:11 PM
Does changing the curve to stepped make any difference?

06-29-2005, 07:41 AM
MonroePoteet - If the object is not moving, and the camera is not moving, there should be no motion blur applied and therefore no transparency issues. But, there is.

Karmacop - We tried to change the curve to stepped and it had no effect.

Called tech support and they seemed to agree that this sounds like a bug.

06-29-2005, 08:07 AM
I don't think this is a bug, but one frame envelope changes are a pain to do with motion blur. Can you try to set both keys to linear and check again?

As far as I remember that would let LW ignore an envelope change for motion blur, at least with LW5.6, I'm not sure if this still works (yup, that was a feature back then, not a bug ! ;) ).

BTW, if this was removed, loads of tricks that rely on motion blurring sub-frame changes to surfaces wouldn't work either...


06-29-2005, 09:24 AM
Motion blur isn't just about motion, it's about the change of an object over time. So as it dissolves over time it "blurs" this change.

06-29-2005, 09:53 AM
To repeat: If you don't like it, put your 100% dissolve at <frame>.1 using fractional frames.

The way motion blur works it to take the number of AA passes (say 5), divide up the time to the next frame into that number of intervals, obtain the state of the object at each of those times, and then combine them into a "blurred" image. Since the dissolve is happening over time between frame N and frame N+1, each of the sub-samples besides the last one is NOT 100% dissolved.

To make it explicitly clear, if you're using 5-pass anti-aliasing (non-zero is required for motion blur) with 100% motion blur, and the dissolve is between frames 30 and 31, then LW will sample the object at somewhere around times 30.0, 30.2, 30.4, 30.6, 30.8, and 31.0 to get the blur.

At 30.0, the object is 0% dissolved. At 30.2, it's 20% dissolved, at 30.4, it's 40%, etc. Only at frame 31.0 is it actually 100% dissolved. By blurring these states over time, you get a semi-transparent object at 31.0.

To repeat: If you don't like it, put your 100% dissolve at frame 30.1. Please don't petition Newtek to change this expected behavior.


P.S. Suppose your object IS moving at frame 30, but is stopped and stable at frame 31. You'd see motion blur on frame 31 because of it's PAST state. Only at frame 32 would you see a non-blurred picture of your stable object. Same idea as the dissolve.

06-29-2005, 12:38 PM
You learn something new everyday! Thanks guys, as usual.

06-29-2005, 10:28 PM
Yep! This seemed as a POV effect case. I agree, this does not look as a bug to me. ;)