Fusion CIS Visualises New 3D Printing Process with RealFlow

The 3D printed object in the film is a complex, spherical cage-like fused-ring structure made of 20 hexagons and 12 pentagons representing a fullerene molecule, called a ‘buckyball’. In Maya, Cinco created previz sequences containing the animated 3D geometry of this buckyball and the camera for all shots above the surface of the resin bath. Fusion CIS, who specialize in physical simulations, created the visuals for the nebula-like resin/laser reactions withRealFlow, Fume FX and Thinkbox Krakatoa - plus the simulation of the resin interacting with the buckyball with RealFlow and Smorganic, and the shading, lighting and rendering in Maxwell Render. Cinco then composited and finished the project, adding lens flares, atmospheric fogs and depth of field blur.

In the first few shots, Fusion CIS devised a nebula-like particle system the show the ‘genesis’ of the chemical reaction using a combination of RealFlow, Fumefx and Krakatoa. Custom RealFlow particle simulations were created as the source chemical reactions, and used as the source for Fumefx sims and processed in Krakatoa. The end results were rendered with Krakatoa.

To gain specific, art-directed control on the motion of the lights at this point, they hand-animated the movement of the leading point of light, a small sphere, and the birth of further light points from the lead light. Then they ran a particle system simulation in RealFlow which added a sparse particle coating on the animated spheres and trails behind these coating particles as dense linear threads. Thesethreadsare tightly-packed lines of particles that are pushed around with gentle forces like noise, and are kept tightly-packed using a particle interpolation algorithm. The threads could also be used as FumeFX sources.

They also created visible lights on the tips of the trails, and a small,splashy burstof particles to start the shot, and instanced small irregular fragments of geometry on these particles with spin controlled using simple expressions. These were rendered with a metallic shader for a sparkle effect.

Two FumeFX sims were created, one for the particles coating the light point spheres, mostly rendered as green, and the other using the trailing threads as sources, which are mainly blue. In 3dsMax, they parented omni lights to the spheres, which cast rays in all directions, and controlled the colour and density of the Krakatoa particles as a function of their age, using a magma flow set-up. This can be done as a post-processing step by including Age and LifeSpan channels and allowed the introduction of magenta to theelements, to follow the client colour palette.

The film then cuts to a wide view of the chemical reaction (below), where more points of light are igniting a complex chain reaction. This swirling mass of light and matter begins to form the 3D printed object.

For the second shot, Fusion CIS created a complex particle system that consisted of 12 elements and more than 300 million particles. The main swirling nebula is made of 10 Krakatoa elements, a murky blue background element and an ambient sparse element of bright, tiny light points.

For the main nebula, they created aRealFlow particle systemsimilar to shot 1, but in this case generated trailing particles for a set of leading points arranged in a circular pattern initially, which then converge to the centre. These thread-like simulations add an intriguing, distinctive look, because they can generate beautiful masses of trails.

Variation was introduced in several ways to give the final compound nebula element a layered, organic feel. From each set of particles coming out of RealFlow, they generated different characters of swirling Fumefx simulations. Some elements were more rapidly moving and dispersing, others slower and more concentrated on their sources. The density slicing settings were varied when generating Krakatoa particles and in the render, different elements were assigned colour variations according to age, based on the approved colour palette.

Moving in toward the centre at the end of the shot, the viewer sees that at the central-point, thereactionis building the fluid into something, accumulating the fluid material that will form the printer’s 3D object.

After this is a shot in which a RealFlow simulation of droplets converge - all at high fps, rendered with shallow depth of field and some Simulens lens flaring from Maxwell Render. When rendering transparent liquids it’s necessary to reflect and refract the environment. Because this liquid is coalescing at the centre of the swirling nebula, they surrounded thedropletwith the nebula by rendering the nebula element from a central camera with a wide field of view. The render was then projected as an animated texture on the inside of a cylinder surrounding the droplet.

In the next shot, the original idea was that the camera should rise up from below the surface, but creating the transition from the subsurface nebular world to the above-surface environment was too time-consuming to achieve within their time frame. Consequently, it was decided to cut to the surface of the liquid and only show the formation of the 3D buckyball. The Fusion CIS team did feel let down about the change in plans, having already produced abeautiful nebula renderand the fluid surface above it, reflecting the nebula.

From this point, the camera moves in close as the buckyballrises from the resin, forming just below the surface in a light-driven interaction.

The challenge here was to show that the buckyball was not just rising out of fluid, but actually forming from an illuminated area just below the surface of the fluid. First, there was no straightforward way to morph the fluid to the shape progressively, as the printer does in reality. Second, achieving the sharp, well-defined edges on the object shape would require a very high-res fluid.

Furthermore, most of the shots showing the liquid-to-object interaction are extreme close-ups and fairly long shots, which means the simulation would have to stand up to close scrutiny. So they simply showed the object rising through a shallow bath of viscous liquid, and made it appear to form from the liquid through shading, lighting and compositing.

The simulation was simple in concept, but not as simple to execute. The complex object shape had a huge number of tiny, upward ‘cupped’ spots for the liquid to get stuck in. This went contrary to the brief, in which any excess fluid was meant to slip off the object slightly above the surface. A scripted forcefield was created, called‘the cleaner’, which applied only to particles touching the geometry and above the liquid surface.

The cleaner pushed particles parallel to their colliding face, in a direction as parallel to straight down as possible. This gave the fluid a lively movement as it flowed off the object that, when strengthened, was actually very similar to motion of the excess resin during the real printing process. To ensure continuity and consistency, the team ran a singlesimulationcovering the whole animation of the rising object. 

To make it look like the object was being created from the liquid, the shader used on both liquid and object was the same, allowing them to blend together as much as possible in the render. In the composite, the edge between the two was also blurred a little to promote this illusion.

The object itself was rendered with a vertical gradient in opacity, so it appears from nothing within the liquid. Close to where the object starts to appear, a copy of the object was rendered, this time with an opacity map set so it is only visible in the area of formation. This copy of the object has alight-emitting materialapplied, creating the internal illumination near the liquid surface. This represents the laser pattern the printer generates to trigger the solidification reaction in the resin.

The challenge of illuminating the liquid internally was the render time. Render times with Maxwell increase by a factor of 2 to 3 if you put lights inside transparent or translucent materials. This was the main reason they chose not to render depth of field blur, instead rendering a depth pass and handling depth of field in compositing.