Dean Grubb tests the new Houdini 10, which offers Motion FX, an enhanced Python and a few more cool features.
After the dramatic UI overhaul and dearth of new features seen by the progression from version 8 to version 9 of Side Effects Houdini, many industry professionals will be keen to see what's in store for its imminent version 10 incarnation. The answer is, in contrast to 9's substantial changes, consolidation and optimization. These efforts focus mainly on improving work flow, refinement of the new dynamics features of 9 and a more solid establishment of Houdini's Python integration via the Houdini Object Module (HOM), with a more complete implementation of commands and more comprehensive documentation.
Workflow Efficiency In Houdini much time is spent wiring and manipulating nodes. Whereas this process can be automated through the use of H-script and now HOM, this requires scripting knowledge that many users may not have. New features to the network editor have been made which will help to address this and make the process more efficient. Central to these features is node behavior akin to Shake. Nodes will now automatically wire together if placed between two other wired nodes and will also disconnect from nodes when shaken. Other enhancements include automatic selection of parent or children nodes when a node is clicked on while simultaneously holding shift or ctrl respectively. These are small tweaks, but with the advent of large networks undoubtedly useful time savers. The introduction of post-its to enable the user to more easily comment their networks and an upgrade to netboxes, which can now be placed inside one another, are also useful additions. Another noteworthy upgrade is an overhaul to the default alt+e expression editor. This has now been bestowed with a plethora of new tools, such as code indenting, designating comments, undo/redo and search and replace. Previously it was not uncommon to use a third party text editor to load when using the alt+e shortcut, this may do away with that necessity. With regard to UI cosmetics, the most obvious addition is the availability of the 'Houdini Dark' color scheme, which is a little easier on the eyes and was a commonly downloaded color scheme file with version 9. The user is even also able to dictate the width of nodes with the HOUDINI_NODE_WIDTH environment variable. Now that's user power!
The Forgotten Context Motion FX is the cool new feature aimed at re-establishing in artists' minds the power of one of Houdini's often overlooked contexts: channel operators or CHOPs. Motion FX is accessible via the context menu on any field in the parameter pane and furnishes the user with a range of animation post-processing options. For example, if the user has key framed a bouncing ball (wildly inventive!) they can then use motion effects to cycle the motion, or add a filter such as smooth or noise. Simply right click, select the corresponding option and Houdini will set up the CHOP network, its imports and exports and some sensible defaults. The user is then free to tweak the effect to their liking. Different animation effects can be layered to quickly create movement of high complexity. It is possible to re-create these effects in earlier versions of Houdini manually within CHOPs, but the point of this feature is accessibility and efficiency. The resulting CHOP network can also of course be examined and edited by the user, massively facilitating the learning curve of post processing with the CHOP's context.
Motion FX are not simply limited to editing key framed movement. These effects can be applied to any keyable parameter. They can be used to post process and filter particle birth rates, gravity settings, modeling parameters and literally hundreds if not thousands of other variables from any context of Houdini. That taken into consideration, this feature is potentially extremely powerful.
The Big Feature One new feature has the potential to make a large impact within the realm of dynamics. This is "distributed simulation." This is the ability to run a simulation over multiple processors and thus on a render farm. This is applicable to simulations based upon fluid dynamics such as fire, smoke, gaseous elements and of course fluids themselves. This, of course, means faster sims. However, there is a benefit of possibly even higher importance. With the ever increasing demand for higher complexity, this gives the artist the ability to run simulations without the inherent memory limitations imposed by using a single render node. Theoretically this would enable simulations requiring an unlimited amount RAM to be completed successfully, provided access to adequate hardware resources. This has the potential to be a major step forward toward the realization of CGI no longer bound by technical prerequisites and of much greater ambition and creativity.
The tracking and sharing of simulation data over multiple render nodes is performed by HQueue, server software provided at no additional cost.
Another new introduction to the dynamics context is the addition of a new RBD solver, the open dynamics engine or ODE. The ODE solver excels at the quick simulation of high numbers of simple shapes such as boxes and spheres. A common complaint of the native Houdini RBD solver is that, although stable and accurate, it can quickly slow down with a high number of individual bodies even with simpler geometry. This inclusion frees the user to tailor their simulations to maximum efficiency, mixing and matching solvers based upon the complexity of the geometry to be solved. It also enables an artist to see an approximate outcome of their RBD setup in minimal time. Then, based upon the success of the less computationally expensive sim, switch to the more accurate native solver, or if need be, tweak the setup further.
There are also numerous other enhancements to dynamics. A new work flow for volume based simulations has been introduced, which allows the user to solve using a low resolution voxel grid, and then re-solve the simulation at a higher resolution using the low-res data as a "head-start." This has the benefit of showing artists a sim preview while simultaneously shortening the time required for the same sim to run with production quality settings. This approach to complex simulation has been implemented in other proprietary software to great success. One example of this being Double Negative, which used this method in their fluid solver Squirt on Harry Potter and the Half-Blood Prince.
Other Cool Stuff The new tools in Houdini 10 are too numerous to mention each one individually, but here are some of the coolest. First, stereoscopic cameras are a lot of fun. These mimic in openGL the three- dimensional effect seen in many recent movies such as Monsters vs. Aliens and Coraline and it is very entertaining to model through a stereoscopic camera with a pair of 3-D glasses on. The new Pyro FX shelf contains some great tools for getting good fire and smoke effects quickly, such as billowy smoke and fireball. These will not produce production ready images on their own but when used in conjunction with the new comprehensive pyro shader and given some refining iterations will output very convincing results. Of course the dynamics networks created by these shelf tools can be edited by the user as they see fit making them as flexible as the user needs them to be. Although the rigid body solver has not seen any significant improvements, cloth, wire and fluid algorithms have been tweaked to improve their speed. Of the numerous miscellaneous shelf tools introduced, two which I find very useful are freeze and knife. Freeze will automatically transfer any object context transforms to the SOP context, and knife acting like an interactive clip SOP, which is much more intuitive.
So What's Wrong? First, although the introduction of the ODE solver will decrease the time to take an RBD sim from concept to completion by allowing faster turnaround of test sims, the native rigid body solver has not been optimized to any significant degree. It remains reliable and accurate, but also slow. This is somewhat disappointing.
The HOM syntax is still too verbose. In many cases, the expression language, and, if required, its combination with Hscript, is still the more efficient option to use in conjunction with parameter fields. I think this convolution of the HOM syntax is one of the major factors impeding its succession.
Finally, a common and quite frustrating problem with the 9.x builds is its frequent tendency to spontaneously cook CHOP networks that are not currently viewed or referenced. This it seems has not been addressed.
Conclusion Houdini 10, while not mimicking the obvious substantial changes seen by 9, is still a positive step forward in the development of Houdini. Features such as distributed simulation, motion effects and the new pyro and cloth tools will benefit artists creativity and save time, while the sum of less hero features such as stereo cameras, new node wiring behaviors and the wealth of new shelf tools such as freeze and knife are greater than their proverbial parts, making Houdini 10 a much more complete tool for the artist. Some niggles are still present, but these are minor and do not take away from this being a very worthwhile release.
Dean Grubb, a graduate from Bournemouth University in the U.K. with an MA in Digital Effects, spent two years in London working for Framestore and Double Negative as an FX TD on various film and commercials projects, including the Oscar-winning Golden Compass and Harry Potter and the Half-Blood Prince. He recently completed work at Digital Domain on the forthcoming 2012 and is currently simulating cloth and fur on Alvin and the Chipmunks: The Squeakquel at Rhythm & Hues.