Faster Fluids A big step toward Houdini's march to be the complete VFX solution for high-end computer graphics was the addition of its own array of fluid simulation tools in the version 9 release a couple of years back. In production, the main hindrance to the widespread use of Houdini's native fluids has been simply a matter of speed. Many visual effects facilities possessed their own faster proprietary fluid solvers or turned to RealFlow. The release of Houdini 11 from Side Effects sees the implementation of the FLIP solver method (fluid implicit particle), which sees a marked increase in simulation speed over other solver types such as SPH (smoothed particle hydro-dynamics). Both FLIP and SPH represent the fluid as particles, but, to evolve the fluid over time the SPH method employs expensive nearest neighbor searches to sample the attributes of the surrounding particles. FLIP solvers avoid this by storing these attributes on a three-dimensional grid, making the sampling of the attributes more efficient and thus resulting in a faster simulation.
This makes Houdini's fluid toolset a much more enticing prospect for production, and also for the amateur artist with professional ambition who wants to tinker around. After all, the simming itself isn't the fun part; it's seeing the results of the sim that is, and this introduction goes a long way to minimizing the wait and maximizing the fun.
The below fluid scene, simulated using the flip solver and consisting of 50,000 particles interacting with four static rigid body objects took less than 10 minutes to solve 240 frames (exactly 10 seconds of film time). This equates to less than 2.5 seconds per frame, enabling the user to experiment with input parameters and quickly see the effect with very impressive turnaround time, a huge benefit for the artist (not to mention to their sanity when the art direction of the shot alters with no warning!).
Automatic Demolition Anyone who has seen any of the summer blockbusters over the last two decades knows that destruction is a staple of a visual effects artists work. Houdini 11 strives to make this facet of its toolset even more robust than it already is with the introduction of automatic dynamic fracturing based upon the commonly used voronoi noise pattern. This tool takes away the need for one of the less rewarding parts of destroying a modelers' lovingly hand-crafted work: pre-fracturing. The traditional method of digital demolition is to, pre-simulation, "break" the intact model and its contents according to predetermined impact points and/or structural weak spots. If the pre-fracturing process is done correctly, the individual sections will be perfectly co-planar so the model appears intact when rendered. Then during simulation applied forces, such as gravity, take their affect and the structure falls and smashes. With this new toolset, the user can simply have their object fall to the ground or get hit by, for example, a meteorite, and on impact, Houdini will analyze the structure and the current parameters of the simulation as inputted by the user and fracture the geometry accordingly and the structure is smashed apart.
This new automatic method of fracturing geometry for rigid body simulations is even wrapped up in a handy little shelf tool named "Make Breakable," which works in conjunction with the "RBD Object" shelf tool to set up your geometry in a dynamics network with fully automatic fracturing within three or four mouse clicks.
Although there is no doubt that the Make Breakable tool will save artists time, my main concern with this approach, and it seems Side Effects would agree with this based on the comments in the tools documentation, is that the amount of user control when using this option is reduced. Yes, this tool does fracture geometry in a physically "correct" manner, but what is correct on many occasions is simply not what the client wants. They will have their own carefully crafted vision for a destruction money-shot that will not be compromised by over analyzing what is physically accurate.
What the client will want is what is most visually arresting. For this reason, despite the undoubted usefulness of this feature for mid and background destruction as the documentation itself recommends, it will probably be relegated to the back seat in favor of the more traditional method of pre-fracturing for hero elements, which is simply more facilitating to art direction. That is not to say, however, that the user has no options when it comes to the auto-fracturing process.
The user is presented with a number of options on the Voronoi Fracture Configure object DOP, the node that is central to the network created by the Make Breakable shelf tool (along with the activation of the Voronoi FractureSolver, which is now contained within the RBD solver). One of the most useful is the "Maximum fractures" option, which allows the user to specify how many times each fractured "piece" will re-fracture on further impact with any other rigid bodies. This can dramatically change the look and feel of a simulation. The user also has control over the points scattered over the geometry surface which determine the density of individual pieces that will be created on shattering, from large chunks to pieces of small fragmented debris. Controls are also available for clustering shattered pieces.
All of these parameters make auto-fracturing more appealing and flexible, but an artist may get into a tricky situation if using auto-fracturing and is then given a specific piece of art direction, which is simply not facilitated by how the object falls or is hit. Overall, the tool is technically very impressive, with the potential to make certain tasks much quicker and easier (such as mid and background destruction as mentioned) and is a very welcome addition to the RBD toolset. For users that wish to forgo the automatic method for more control, the "Shatter" shelf tool, introduced in Houdini 10, employs the Voronoi Fracture sop and is a very capable and fast method of pre-fracture. The use of this tool allows the user to explicitly control the positioning and bias of fracture points using any method they wish from the extensive SOP (surface operators) tool-set. The best of both worlds, one might say.
Houdini 11 sees the switch to Python 2.6 from 2.5. Users now also have the option to choose which Python distribution they prefer, which will be a nice addition for Houdini's power Python/HOM users (this includes Hython, the command line implementation). Houdini now will also attempt to match the currently installed system distribution of Python, so if a user has 2.5 installed, Houdini will use 2.5, if 2.6 is installed it will switch to 2.6. If the user has no version installed, Houdini will use the 2.6 distribution, which it ships with. There are no specific changes to how Python is implemented, however, meaning on many occasions node expressions in Houdini's native H-script and/or the expression language still execute faster.
Lights, Camera, Render!
With regard to lighting and rendering, Houdini 11 has concentrated on the improvement of existing tools, but has also introduced one or two new ones. First, there are substantial improvements to the handy environment light. This now gives overall better performance than the already impressive render times seen in Houdini 10. The environment light also has new features including a "portal light," which simulates environment lightning from a limited source, like a through a window, and sky lighting with physically correct time-dependant positioning of the sun. Another welcome improvement is support for multiple environment lights in the same scene, which correctly interact with regard to luminance accumulation and shadow interaction.
One of the new implementations mentioned to lighting comes in the form of the geometry light. This will cast light from geometry in accordance with the objects shader parameters (for example, the color of the emitted light is picked up from the shader color). What is most fun about this tool is that it supports animated geometry, including geometry animated dynamically through a simulation, which can make for some very interesting lighting effects. One drawback is that the geometry light can become computationally expensive, but used astutely this is a very interesting and potentially useful tool.
Other improvements to lighting include performance enhancements for area lights, a handy ready set-up three point light rig to simulate the traditional key, fill and bounce scenario and the new indirect light, which gathers the functionality of many previous tools centering around indirect illumination, such as GI, caustic effects and photon mapping all in one place.
After IPR (Interactive Preview Rendering) mysteriously disappeared from Houdini 9, it had a welcome re-introduction in 10 and has now been upgraded. IPR can now be network distributed and is also compatible with the useful region render tool.
Other miscellaneous improvements include new CHOPs centered on crowd simulation, support for Disney's Ptex format, a range of new fur styling tools and an impressive array of new volume manipulation sop level tools. One last note, there is a new "lazy-line" option for node wiring, which I found so amusing I thought it worth a mention!
In conclusion, Houdini 11 sees some very noteworthy improvements. FLIP solver fluids will make a huge difference in the use of Houdini based fluids in production and is a major enhancement to Houdini's fluid arsenal. Efficiency and ease of use enhancements to lighting tools and the RBD toolset will also enable faster turnaround of common VFX tasks. One downside to the release is the lack of any real brand new "hero" tools, but the introductions to augment existing toolsets are excellent.
Alexander Carson-Brown is a long time Houdini user and a graduate from Bournemouth University’s MA Digital Effects programme in the UK. He is now an FX-TD in the motion picture industry and has produced effects work for numerous commercial spots and many films, including Harry Potter and the Half Blood Prince, Alvin and the Chipmunks 2, Wanted and The Golden Compass.