Steven Knipping discussed the opportunities Houdini gives the artists and its development and also presented his useful Applied Houdini courses.
Hi, my name is Steven Knipping and I do the Applied Houdini tutorials! I was born in New York, raised in NJ, went to school in Philadelphia, and have been in California ever since. I was fortunate to be recruited by Dreamworks Animation in grad school and was even luckier to work on How To Train Your Dragon as my first film. Though I thought I wanted to be a modeler at first, the problem-solving aspects of FX ended up interesting me more, and I had my first opportunity to do that on The Lone Ranger at Atomic Fiction. After a brief stint at The Mill in NYC, I was back at Atomic Fiction, and then now finally ILM for the past 4 years or so! Along the way, I’ve worked on over 20 films, including most of the new Star Wars and Avengers movies! You can check out my IMDB here for a list that is hopefully up to date.
I get up to a lot of projects outside of work, too. Past favorites have included a long stint of bronze metal casting, building a pirate-themed bar/lounge, and currently working on a 16-foot traditional wooden sailboat! I hope to have that done soon. I would also highly recommend archery. Basically anything traditional/involving your hands to complement being on a computer all day.
I think what appeals to me about Houdini is how granular it is, and the comfort in knowing that no matter what is asked of me, I can always engineer my way out of a problem because I have total access to everything (geometry data, simulation data, low-level tools) all the time. Being able to combine small bits of powerful functionality into a node, and combine those nodes into a system, and combine those systems into an ultra-simulation gives rise to complicated effects that are simply not possible otherwise.
Plus new features are constantly being added to it! I’ve been using it since Houdini 10 and there’s been so much improvement on what was already a great platform even then. There’s so much in there that you can’t possibly know every nook and cranny, especially as they are adding more things faster than you can learn them – which is a great thing! It’s been said that it is a lot like visual programming, – you are connecting little modules together, – but I think even people who are not programmers can still see how to combine and recombine nodes and tools made by others and find an intuitive and powerful way to get to a very specific outcome.
With many programs, you are relying on what comes with the program to make your effect, whether it’s a plug-in or built-in, but then you just have a “black box” that you can’t access. With Houdini, you’re given that box, and then you can decide whether to use it as it is or just immediately open and change it to do exactly what you need. It will look better and will be faster in the long term if you can always do exactly what you want rather than fight against the limitations of how a developer happened to design a tool.
Introducing Houdini and Procedural Approach
Like many people, I used other software solutions apart from Houdini and started using Maya for everything back when I first started. These days, I still use Maya for certain tasks – mostly organic models that don’t lend themselves to being built procedurally. If you’re reading this interview without really knowing about Houdini, know that it positions itself very strongly as a procedural workflow that is fundamentally different from other software solutions. Proceduralism, in this case, means that one thing leads the next, leads to the next, and it’s like a snowball rolling down a hill getting built up along the way.
To put it another way, you have one input that feeds into a machine (the “machine” that you built in Houdini) and then your effect comes out the other side. You are building a machine that takes an input and gives you some crazy output which you then modify with another machine. Those two machines together could be a system which can also become a part of another system, and so on. This allows you to build wildly complicated effects that you could never do with approaches where every click you do in a program equals 1 action. Instead, you are designing a process to make something complicated happen – click by click or keystroke by keystroke yourself. The benefits of this approach are further compounded by the fact that in production your input (character models, character animation, terrain) is being revised all the time. Instead of having to manually redo work, you often can simply rerun the system again with the new input, and you’ll get your new output. Go on and go home early, Houdini will chew on it overnight for you.
As for the best way to get an introduction, honestly, as the creator of Applied Houdini, I would strongly suggest Applied Houdini! Together with downloading Houdini’s free version (completely free, and with the exception of lower image resolution and a watermark, it is fairly unrestricted!) it is a great introduction to the program. I offer several free lessons on www.appliedhoudini.com that are aimed at beginners, and many people seem to like them. I hope you will too!
Houdini and Game Production
Procedurally generated content in games dovetails quite nicely with Houdini’s approach. Being able to generate hundreds of variations of a vehicle, building, or any other assets based off some rules with varying inputs is a good example of how this workflow sees use in games. Tools that automate all kinds of things with many intricate pieces that can be controlled through larger systems (stairs, ladders, trees, brick faces of buildings, crazily complicated robots) are perfect candidates as well.
I am also aware of the Houdini Game Development Tools toolset, which is available now. Unfortunately, I don’t have any personal experience with it, as I do most of my work on the offline/VFX/movie side. Still, as I feel there are more Houdini-for-videogame tutorials coming out these days rather than for VFX, it looks like people really seem to be taking advantage of it.
Volumetrics: Explosions & Volumetric Clouds
Explosions are always a super fun thing to do in Houdini and in visual effects in general. Houdini has been a great way to do them for a long time, but of course, it really comes down to the artist to make them look truly nice. Houdini’s workflow and the Pyro solver, in general, are super great and have only gotten better over the years.
In order to learn how to do it well, once again you really just need to download Houdini. I’ll just have to plug Applied Houdini again as well – the Dynamics series teaches how to simulate and render explosions, and the first lesson is free. It can be daunting tackling explosions for the first time, but relying on the advice and experience of somebody else to help you get started will go a long way to making it easier than you might expect!
Volumetric clouds specifically have their own Clouds tools, but rely entirely on understanding how volumetrics work in general. I’m not sure video games at the time of this writing are really ready for real-time ray traced volumes, as they are extremely expensive computationally, but as GPU technology improves, I’m sure we’ll get there sooner than later.
Rigid Body Dynamics
Just like with volumetrics above, this is another area that is extremely computationally expensive and therefore very difficult to do in real-time. The technology to compute rigid body dynamics as well as the hardware in terms of the GPU is getting better all the time though, and even now there are relatively simple rigid bodies in video games.
Considering that simulations I work with can involve 10,000’s of pieces all constrained to each other with complicated forces to simulate bending, glue, twisting, etc., and that those sims can take 15-20 seconds to frame, I don’t think we’ll be seeing anything of that caliber in real-time soon. I don’t honestly know how developers can trick the game engines to get around these restrictions but are probably working around the clock to figure it out. I think we’ll see more complex sims in games soon, but our expectations will have to be reasonable until the hardware improves.
Learning Houdini and Its Most Interesting Tools
Houdini’s liquid simulations with the FLIP solver and the grains solver (for sand and snow) are currently a kind of the frontier, in terms of people trying to wrangle out as much detail as possible given the constraints of the computing power we have today. The general approach that people apply to those and to learning Houdini, on the whole, is mostly just using it and working on projects as much as possible. Work on projects, show them to people to get some honest critical feedback, and improve in the next project. If you get stuck, ask someone for help! For most people that will probably be a post on the forum, but sometimes that’s all you need to get past a roadblock.
There are definitely no books that I am aware of that are up to date in this field. SideFX has some introductory tutorials that are good. If you are already familiar with Houdini, I would also recommend the short lessons that Entagma puts out: they cover more specific/obscure corners of Houdini. And of course, once again – Applied Houdini.
What Should We Expect in the Future?
The major releases always add lots of cool new features and incremental updates on existing features (speedups to existing solvers, new tools that consolidate common tasks into one friendlier node from many). I’m not entirely sure what the new version will have, but we won’t have to wait for long now!