Hummingbird Wings Motion Blur?

[primarysmirk]

Hi all. Would love to post this elsewhere, but I'm a student without access to license info. My thesis film involves a cg origami hummingbird that is working perfectly, except for the motion blur on the wings. Hummingbird wings beat at an incredibly fast rate and are essentially just arc shaped motion blur. I haven't been able to match the level of motion blur, and was wondering if anybody has ideas for ways to approach the issue. Here's what I have tried, and why they haven't worked.

- Velocity pass used for vector blur in Nuke. This seems to only calculate blur for the current frame based on the immediate surrounding frames, not the whole flapping cycle. The end result in Nuke is that each frame is blurry, but the blur is only a fraction of the whole arc. It also introduces occluded geometry issues.
- VRay physical camera with the shutter wide open at 360 degrees. Again, the blur does not span the entire arc.
- Regular camera with motion blur turned on. I have tried many combinations of durations and geometry samples. It seems logical to me that the ideal solution would involve a very high duration of frames and geometry samples to account for blur across the entire flapping cycle. Unfortunately as these settings are turned up the render times skyrocket to impractical levels, and that's with subdivisions and screen resolution at an unacceptably low level.

Something else that I have been playing with is frame rate. Hummingbird wings beat many more times per second than is possible in 24 fps. At speeds of around 60 fps I get an extremely nice smooth animation of the wings in all the necessary positions (as opposed to just up, middle, down keyframes). Exporting at that rate isn't really an option since everything has to be comped into 24 fps footage, so most of the frames would be thrown out anyway. One thought I had is to use the high frame rate to achieve very nice blur overall and then just export enough frames to be 24 fps (every 3rd frame or every 4th frame etc). Again, this runs into the issue that the duration of frames would have to be that much higher to account for the high frame rate.

Last note. I have read about people using multiple wings in different positions to achieve blur. The way the rig is set up is that as the wings move the entire body of the hummingbird is getting properly pushed and pulled through a series of IK chains. I haven't yet thought of a way to modify it to account for more geometry.

Thanks for any and all suggestions, and congrats on making it through an essay of a post!
–P

[joconnell]

it's similar with helicopter rotor blades. You've got to use a cycle as you say, but once you've set your up and down keyframes, you've got to scale them in a huge amount - you're going in to sub frame animation where you have multiple cycles of the up and down animation withing the one frame - in 3ds max it's called "ticks" (Minutes : Seconds : Frames : Ticks) and then you're also going to have to use enough samples subdivisions on your motion blur to capture it. I've done this successfully before where I've done light pen animations - having a very bright object moving along the full length of a path which spells a word - you need lots of geometry samples to be able to have enough time samples to get a nice smooth arc, and then lots of subdivs to clean up the noise (or high aa).

[primarysmirk]

It would appear that I have it working using a combination of techniques. I did have to modify the geometry and rig to accommodate extra wings. I also, as you suggested joconnell, scaled in my keyframes. I keyframed one wing flap over however many keyframes it took to look nice, then scaled it in so that there are about 7.5 flaps per second. I found that the best motion blur settings were a duration of just under two flaps worth of frames and geometry samples of approximately twice that value. As for the render time, it was an extremely simple solution. I was smoothing all geometry using viewport subdivision and simply unsmoothed the wings since they are so blurred. Problem solved- the thesis continues!
-P