Jump to content


  • Content count

  • Donations

    0.00 CAD 
  • Joined

  • Last visited

Community Reputation

0 Neutral

About kurtw

  • Rank

Contact Methods

  • Website URL

Personal Information

  • Name
  • Location
    Los Angeles, CA
  1. SDF Blending

    Does the Volume VOP not support VDB for instance? Actually the supplied VOP will work with VDB's but VDB's handles it's data a bit differently, where the range where it will work is .. very small depending on you scale & grid resolution. The solution is to ensure you have enough "working space" around the VDB surface by adding enough padding around it. To do this during the VDB from polygons step, make sure you raise the Exterior Band Voxels so it encompasses the areas where you wish to blend/fillet the vdb's together. The default of 3 won't work, so I found values of 20-100 would work, it will slow things down since you are creating a larger "band" of active voxels around each of your inputs, but it'll still be an order of magnitude faster and more precise than using iso offset. So for very irregular shapes, you'll want to put down an Open VDB Visualize or VDB visualize tree and look at the active voxels of both inputs to make sure the working active voxel radius will give you enough working area for the surface blending. One thing I've tried and I've seen others tried is to manually [as a post process to the vdb surface generation] enlarge the active VDB grid, either by combining the two input grids into one via a volume mix, or using vdb resample/reshape to make a bigger grid, but that doesn't seem to work with the volume vop as presented, a step is missing that is applied during the active voxel generation of the initial surface creation. Either way, adding enough active voxel padding will work.
  2. Which GPU to choose

    ATI/AMD's for me historically has been nothing more than a wet fart driver wise, especially with linux.
  3. Titan X and 64bit OpenCl computation

    Thats interesting I was under the impression I'd have to track down a used Titan GTX/first generation board for the FP64 support. So you can get away with openCL pyro/flip on a TitanX/Z without any performance loss? That's a huge savings over the M6000.
  4. I'm with everyone else, I'd go with more memory followed by the most CPU cores you can afford. After that comes GPU and and SSD is pretty low on the list. Core's a big thing now I'd go with as many cpu cores you can afford, a hex core would be nice, or a octo core if you can swing the cost on the newer Haswell-E i7's. I'd do that even if It meant going with an older GPU for a bit. I've seen quadro 2000 boards go for $100 or less on craigslist, that's an option as a stop gap until you can upgrade the GPU later.
  5. Commence the jigglin!

  6. Constraint FEM

    I've used soft body constraints in H13, SBD set to soft. I'll generate a selection volume for my tet meshes and use that to generate a softbody point constraint group. In Dops on the SBD constraint node you can a pointlist function to extract the group's contents into the node. eg. 'pointlist("../../group1","groupName")` I'm using a static tet mesh which is being deformed by a point lattice operation as the simulation input. That way I get a animated "rest state" for my FEM that is stable. Make sure you are importing the rest and target geometry so that FEM knows its dealing with an animated tetmesh, then you'll get something that SBD constraint will work with. After that it's playing with the values, number of nets, etc to get something that works nicely.
  7. Textures not updating in viewport

    I was wondering about this. Running 12.1.230, and I'm trying to force Houdini to update textures per frame in a shop. I'm doing game engine visualization using the openGL rop and the results are inconsistent. I'm animating a bunch of prerendered image sequences on sprites and having a heck of a time getting it working. I tried adding opcook and glcache clear to my pre frame scripts but that results in a core dump. The weird thing is, it sometimes works. I close and reopen the file and it doesn't work, just renders the first image in the texture sequence. /edit Render> Update textures just crashes my session. /edit #2 Trying it in 12.5 the behavior is more predictable. I can use opcook -F to force the shaders to update on each frame and that gives me a workaround for now
  8. Houdini 13 Wishlist

    Better hardware/display support for game development. This is the classic scenario I'm bumping into. Major game studio shows an interest in using houdini vs max & maya. I get absolutely burned trying present things in Houdini, and as a result have to use Max and Maya to output/demo things. Sprite display in opengl seems wildly inconsistent while it "just works" in the other packages. I mean its almost there, until you display, say a bunch of pyro/flip renders onto cards for sprite particle animation for real time view port playback, and then you get weird inconsistent stencil buffer issues, odd composting/clipping with a skybox turned on, etc. So then it begs the question.. graphics card issue? I'm playing whack a mole on the hardware side with everything from a quadro 4000 and fire gls to franken pc geforce cards. It would be neat to have all the hardware rendering/display options on one universal configuration pane somewhere, ie. what viewport drawing to use, texture filtering, etc Instead of jumping around to 2 and 3 different panes. I guess I could write a otl that toggles those...
  9. Videocard for Houdini

    I was thinking of getting a titan, which concentrates a lot of processing power on one gpu vs some of the higher end geforce cards which is basically 2 gpus SLI'ed on one card. I'm not sure if Houdini's Open CL can take advantage of multiple cards? I just upgraded my system to a 6 core I7, 64GB of memory and ssd drives dedicated to cache files and system swap, and now my geforce GTX 470 is a bit dated, its faster for me to run my simulations on cpu vs open cl. So my big limitation is GPU, and the first choice seems to be the new Quadro K4000... But the Titan has 6GB of ram and is within the price point and I'm thinking I'd like to have the extra memory to run a larger simulation grid. It may be worth the tradeoff going with a ultra high end "gamer" card... I haven't had issues running with geforce at home [i've been running Quadro 4000-6000 series at work for the last while]
  10. Freelance daily rate

    Regarding vfx wages, a lot of it comes from visa data. For us based jobs, anyone who has worked in the US under a work visa at any company has their salary and job title accessible by a freedom of information request [other than that and the visa start date, the data is anonymous]. Many of the spreadsheets out there use data from this, A number of websites have search engines to look the data up. www.h1bwage.com is one, but the problem with work visa data is your job title will vary wildly. At one studio you will be an animator, at another a technical director, at another a digital artist, etc. So you can search for a few job titles [ fx animator, effects animator, technical director, animator, digital artist, graphic designer, etc] and collate the data to get a feel for what a few rates are. Note that work visa rates are generally medium to long term contract or staff work. www.glassdoor.com has a few facilities listed on there, but that one is fed by user survey data. The Animation Guild has a salary survey as well as contracted minimum rates for job categories, but thats mostly data from Disney/Dreamworks feature animation, and a number of smaller unionized animation studios [Nick, Cartoon Network, etc]
  11. debris creation... how you doo

    It depends, but most of the debris I've done on shows is a combination of modeled objects driven by particle motion and RBD's, combined with some particle elements [small bits of secondary particulate motion] and some volumetric dust layered in. On one show we tried reusing the parts library from a sci fi show that was in production at the same time, we got busted because it wasn't period correct for the film I was on. Production green lit us to run over to turbosquid and we picked out some junk we could throw around in our scene and the problem was solved. Now if its debris destruction of a prop/set piece, then there is some more modelling work, ranging from model swapping and having some hero sections that are broken apart then adding some procedural chunks created by voronoi with noise, etc. Problem is you can often spend a fair bit of time trying to make voronoi .. look.. like not voronoi but something more natural looking. I'd like to see a FEA setup for houdini, FMM for Maya/Max by pixelux is a handy toolset at an amazing price [$299] I've been playing with that in a demo version, Not used to working with maya these days but I guess one could write a otl that connects to maya with the FMM plugin via python Hmmm.
  12. App For City Modeling

    Looks like Sim City after it got hit with the ugly stick.