by Alexander Weide 2016 April 07

A few days ago I was looking for lava footage and reflecting on how far I could take a lava simulation. Lava is not the easiest material to work with and my aim was to produce a realistic effect. My research eventually led me to a Vimeo post by Graham Collier, in which he shared his lava asset. Thank you for that, Graham!

I looked at Graham’s test shot, started to play with the asset, learned a lot in the process and eventually decided to build upon it. I experimented with temperatures and heat dispersion between the air and the cold and warm particles. And I decided to set an aggressive schedule for this project, even though this was new territory for me. I therefore needed a way to simulate quickly so that I would have sufficient learning and experimenting time.

GridMarkets supports Houdini simulations and I have used them before.  Even though the service is in beta, my GridMarkets experience was excellent - and importantly the service is very fast. During the time of my research for this project, I learned that GridMarkets had released a new iteration to their HDA. As the previous GridMarkets HDA was well done, I was anxious to give the new release a try.

But during the beta period, GridMarkets has restricted the sizes of user data sets. (I’ve been told that once out of beta, this restriction will be lifted.)  So, I needed to figure out how to deal with this constraint and still get my sim done. This was a challenge as my sim needed over 300 gb of disk space.

So I started the simulation at my workstation. Later I found the new Houdini feature called simulation checkpoints - which gives you the possibility of creating initial starting points on the fly every 5 frames or as specified. That could be a nice way of reducing the disc amount required on the GridMarkets cloud, because it would be possible to simulate the entire flow, but have only every third frame saved, thereby reducing the data amount. (For my next attempt at this project if the GridMarkets beta restriction is still in effect, I plan to calculate the frames in between locally.)

My workstation is very fast and Linux based. Linux Mint 17.3 is the best OS that I have ever had - fast and very user friendly. It was easy to start the simulation process on the console and I saved a lot of time and computer resources. I recommend to use Linux for every Houdini user. In Linux I am able to connect the CPU, the opencl gpu from my nvidia card and the gpu from inside my i5 together.

After 6.5 hours, the simulation was complete and my working hard drive had a lot of data.

I made a playblast from it and uploaded it to Vimeo.

A day later, I realized that there were a lot of people out there who were very interested in that simulation. So I decided to build a nice looking VFX shot with smoke, heat flickering and background. This was also a good opportunity to build a steam system emitter based on heat! I could use the heat pass in compositing to blur and transform the image to get a nice heat flicker effect.

When the the heat and smoke simulation was complete, I decided to take out my photogrammetry 3d scans from my floral foam experiments a couple of months ago.

Floral foam is very nice for sculpting rocks in 5 minutes, because you can carve into the rocks with a knife and break out some parts. The resulting texture is similar to real rocks.

Then I made a lowres version of the simulation sequence to get it worked as a geometry light source and build the layer setup. I uploaded the entire geolight bgeo sequence to GridMarkets via FTP and started the process. The data amount was quite low: 1.4 Gb of data.  A mere seven (USD) dollars and 1.5 hours later all 400 frames of the backround pass were ready for download. It was faster than I expected and GridMarkets’ new closed-beta HDA is far more advanced than the current public one. It will be a magic moment when it comes out for all the users and especially the Houdini Indie users.

At the same time I started the rendering progress on my workstation and yes, it took quite a while, but I rendered only the lava fluid mesh. One frame took between 1.5 minutes to 4 minutes in FullHD. It was long compared to the 50 seconds per frame on GridMarkets, but okay... The upload of 326 GBs of fluid data was not an option. It would have taken more than 2 weeks, so I think it was a good decision. On Thursday I started the compositing progress, and combined every element: the background, the lava layer and all the compositional elements. Some of the elements were photos taken from asphalt streets around my house or from the Dresden flood in 2002, which destroyed entire streets of our city. It’s incredible what water can do to asphalt streets - it bends and deforms the asphalt so that it takes the appearance of cold lavastone.

The stones in the background are projection painted stone textures. I used the projection paintings to give a better depth to the stones.

I did not render deep data on the shot. Every element was composed in a traditional way as a combination of Nuke and Blender and a bit of relighting with Houdini’s compositing tools. The relighting was only one extra pass to make it much more believable.

The cut of the Vimeo short was done mostly in Blender and later the export in Handbrake.

The entire shot (including the master video file, which is upsized to 4K) and all sub steps of cache files took around 396 GB´s of disc space.

I worked an average of 12 hours a day, because I wanted to be ready by Friday, my deadline. I got the job done on Friday at 11.43 pm Berlin time. It was very, very hard because I had to manage all the simulation tests and deal with everyday life in between, but I know now how far and fast I can go. And I believe Houdini and GridMarkets made me faster than all my local competitors. Maybe in the future I can go even further since I significantly updated my system: SSDs from OCZ ( I think they are the best on the market) and SSHDs from Samsung - and let’s not forget my operating system, Linux Mint 17.3 Rosa. I am already thinking about my next project, an emotionally charged music video with stunning visual effects, so stay tuned...