Twister
I’ve received several requests for information in how the tornado in “TWISTED” was executed, so rather than try to answer all the E-mail posts, I’ve created this page as a FAQ of sorts to answer these requests. I’ve also unfortunately lost several requests in a recent mail disaster, so if you’ve asked and haven’t received a reply, I apologize. Hopefully the information on this page will make up for the delay.
I started with stacked particle emitters spraying particles around spherical wind warp objects. Using a negative wind strength value, I saw that the particles could be held in a circular pattern. I used 8 or 9 sets of these for the narrow portion of the funnel. Ten more were used for the wide upper portion. Where possible, I arranged the emitters in adjacent pairs at 180 degrees from each other. This is to hide the particle origin and make the particle flow appear continuous. For a more varied particle flow, I linked the spray emitters to the wind objects both above and below the nearest wind objects in addition to the adjacent warp. This forces the particles to take a wider vertical path around the wind objects, helping to eliminate the “hoop” effect. It also helps to use a reasonable “variation” setting. The wind strength and particle speeds must be reconciled or the particles will travel in an elliptical path. The higher the particle speed, the stronger the wind strength.
The size and type of particle is also important. I used the facing particle type, and picked a size that would close the open spaces between particles where I wanted the funnel to appear opaque. These places are at the top and bottom of the funnel. Down the narrower portion of the funnel, this was not a big worry since I decided to use smaller, fewer particles in this area, exposing the textured, lofted funnel body.
At the base of my tornado is a dust cloud which is formed in the same way, using a couple of different material colors. I also used the triangular facing type to add fury and texture at the core. I used six emitters with a particle count varying from 200-350 particles each. Here is where the higher number of particles were used. Finally, as if this was rendering much to quickly, motion blur was applied to the particle emitters. This makes the moving particles appear longer and softer. It also greatly lengthens rendering time. For compositing, a ground plane with a matte/shadow object was placed to receive shadows.
The material for the particles is an important step to creating the soft cloudy effect. I used an opacity map that is essentially a radial gradient. This produces a soft-edged round shape. One key to this is to use the face-map feature which centers the opacity map on the particle. Keep in mind that rendering time is adversely affected by using an opacity map in this material. You may wish to test the rendering time and appearance of the particles without the opacity map and decide whether or not to use it. The material for the funnel consists of an animated texture and bump map.
The silo was modelled in pieces which were animated and attached to a noise space warp to add a waving effect as the twister hits. Since my background plate was a still photo, I added some foliage objects in the foreground. I then bound them to a wave warp object which had an animated noise transform attached. I animated the amplitude values on the wave warp so that the effect of the wind blowing the plants would gradually increase, corresponding with the approach of the twister.The funnel animation was accomplished by morphing, moving and rotating the main funnel loft. The emitters were linked to the wind objects which were keyed to mach the position of the loft object every ten frames. As the funnel bends and twists, the attitude of each emitter is animated to maintain a parallel position with the funnel.
In the end, I rendered the animation in two layers. First the upper clouds, and funnel object along with the silo and foreground plants. In the second layer, I rendered the outer particles of the narrow main funnel midsection, and the lower debris cloud. I used matte materials in the second pass which were applied to the objects rendered in the first rendering pass (with the exception of the particles in the upper cloud), to assure a good final composite. The frames from the first pass were used as a background for the second layer. The rendering took ten days on a 200mhz Pentium Pro with 128mb RAM. The bulk of the rendering time was used to render the final 200 frames where the twister nears the camera and gradually fills the screen as it hits the silo, and leaves the screen. The entire animation is 650 frames long. It was originally rendered for PAR output at 752×480.Obviously it has been done to perfection by ILM in TWISTER, using the ubiquitous “in house software” solution that the big houses use. What I set out to do was see what I could push MAX to do without using a 3rd party plugin. I learned quite a bit by working through it and was very happy with the result.
