As you all know, I have been working on some animations at work, made me think back when I used to work in broadcast advertising and motion graphics at Channel Nine.
Back in early 90’s the Head of Communications from Sunrice approached me to collaborate on a commercial project requiring rice to do some, well…unusual things.
I saw an ad in the states, of photo-realistic 3D water. It would leap into the air to form shape after shape: first a brain, then a bridge, then bananas. After the bananas form, all but one splash to the floor; the last one then peels itself and half of the peeled banana splashes to the floor. The creative ideal was for the water to morph between shapes with a vigorous life of its own. When it formed the shapes, they needed to be detailed and sharply defined, while still looking like water, not like geometry with a water shader applied.
So, when I was asked to do a similar approach but with rice as the subject matter, the TVC concept was both incredibly exciting and remarkably challenging! Morphing is of course a very cool effect and, by pure coincidence, I had just finished some R&D on morphing two weeks before I got the contract.
My R&D at that point was that I could morph 2D shapes into any geometry, but that was all. The challenge now was to make it morph from shape to shape, to clearly define the complex forms, and to do so with a lot of creative control and in 3D. Based on my early Softimage 3D morph target based work, I was optimistic about this, however, speed and volumetrics was the main issue. My early tests were quite slow with a lot of calculations involved, as I needed to do a lot of high level maths and physics. I sped up the simulations at the end considerably for “Whole”, because the main shot of the project was 650 frames of continuous rice action – and it had to be a single simulation.
They would decide how they wanted the water to look and behave at any given point, and I would go away and write a piece of code in Softimage that would give me the controls we needed to achieve what we wanted, I exchanged Quicktime movies and test renders via ftp daily to show the progress.
The behavior evolved dramatically over time as the possibilities I found were explored. In the end I achieved virtually everything that the client wanted, and almost everything that was technically feasible for the current capabilities of Softimage. In Softimage there are a couple ways I was able to do this but it was very difficult to get the morphed rice to sharply define any sort of detailed geometry, tending instead to make a sort of globby approximation to the geo. Finally, it’s very difficult to control the behavior as the forms the shape. Typically, particles do a lot of unattractive orbiting and circulating around parts of the geo rather than settling down. Apart from the obvious challenges of jumping shape to shape and controlling how the transitions happen, this has the big downside of communicating to the viewer exactly what it is doing. If you look carefully at the ad, you can see the stray particles colliding with unseen barriers, and forming static, sharp edges. After that, it’s a matter of selecting the target sites on the geo for each rice particle, defined in terms of a kind of UV map data structure I created that’s independent of the object’s position, scale, rotation and deformation (needed so the object can be animated). To give us flexibility in how the transitions happened, I built in 3 ways for particles to select targets: nearest poly with available space, farthest poly with space, or random poly with space.
The next step to getting this to work was to design naturalistic forces that would pull the particles to their targets in a controllable and attractive way. A simple-minded force of attraction from the particle directly toward its target works, but is ugly. Particles get pulled on more or less straight-line paths like little robots, then overshoot their targets and start orbiting in toward their targets. A small amount of this orbiting is cool, as it gives the overall behavior a nice organic feel, but it’s hard to control and tends to produce groups of particles that spin endlessly like little buzzing bees. To get more control, we designed a targeting force that included a pull toward target, a counter-orbiting force, and a radial variation on the force components so the fluid would get drawn horizontally until it was beneath the targeted object, and then get pulled vertically. This setup gave me control of the degree of clumping, and also allowed me to ‘sculpt’ the shape of the incoming rice flux and make it more funnel-shaped, which was especially useful for the transitions.
Although I rarely get the luxury of time to do this kind of R&D during projects, as important as time was the level of collaboration and communication between everyone involved. Between the guys at Sunrice and I, there was almost continuous communication and exchange of ideas, plus the very important but unusual admittance of where there were technical problems and limitations. I like to tell clients “anything you can imagine, I can do” and while I can do achieve great things in advertising, we all know this is quite optimistic. In the end, I pushed the technology I knew (Softimage) to a new high, and also created the best possible product for the time and money.