narbuckl

I have had experience using 32+ core Xeon machines vs. using a hexacore core i7 that is overclocked and from what I've experienced the overclocked i7 was better for workng with large data sets in real time and keeping things interactive, however when you look at tasks such as long renders especially when using tiled rendering, the ability to delegate to 32+ cores is very helpful. The other thing that a high core count xeon will do is allow you to be able to run multiple background tasks without much noticeable affect on your open scene. Personally I enjoyed the realtime horsepower and data chewing capability of a single overclocked processor, but if you need to do lots of tasks at once while also working the multi processor xeon's will do a better job. Another difference I have experienced is high end gaming ram vs ECC server ram. Almost any mobo for xeons will want the ECC ram which is stable and a good work horse, however holy crap when I put in 64GB of Quad Buffered 2400 I was able to work extremely fast with huge geometry. It really depends on what other resources you have I think. Just my two cents -Nathan

Great resource on matrices and quaternions: http://www.tokeru.com/cgwiki/?title=HoudiniVex

If three is some part of the setup I can shed light on I would be happy to, just ask It gets a little involved simply because I really wanted a very specific behavior of the lightning, and I wanted it to be directable. So everything is done to cater to that purpose. If the scene complains about missing asset definitions I can assure you that they aren't necessary for the fx network to compute. It's probably just the backdrop and some informational nodes. The backdrop was modeled after a standard photoshoot type backdrop so L shaped with curved intersection. -N

Hey looks like the upload is working again. Enjoy. You may need to swap out some custom variables for file path names. -Nathan TEST_FX_twcFX_001_001_restore.rar

Found the file but odforce is having some issues with posting files on their forum, shoot me your emails, otherwise I will post it once Marc fixes the upload file problem. -N

Thanks, glad you like it ! And yes Houdini is a magical creature! One of the things I tell Junior TD's or students when they start is that you need to treat Houdini like a blank canvas with all the paint brushes in the world. You can literally do whatever you want, but its not going to tell you how to do anything, don't be afraid to just try shit based on the 'this should work' logic, because thats all it is Don't be afraid to break it! Its all just data and figuring out different ways to get from point A to point B with that data. That test was from an old job directory, I'll see if I can get my hipfile unarchived for it and let ya have a go if ya want! I know I always learn the most by working backwards through more complicated setups, plus you get to see how other people think! If I can't get it onlined I'll put together something new for ya to play with, if thats something you might be interested in? -Nathan

Hey there, an easy way to get gradients on curves is to use fit with the point number with a max of npt (total number of points)-1, you can also use the uv texture node set to rows & columns which should produce a 0 to 1 u value gradient. In terms of lsystems, unless you are customizing it a great deal, I would turn to pops to create your lightning arcs. More control and organic behavior possible there. Here is an example of a 'Rolling lightning' test I did for a job way back when: For this I had 'arc points' scattered in the volume and a single particle would lead the direction of the arc. As it moved across the point cloud it would send out replicants that would try to connect nearby arc points as well as respect any custom forces (for example I wanted them to go from top to bottom, so they would do a point cloud lookup and get their desired target point then add in whatever custom downwards vector or noise I specified). If this seems like overkill then stick to lsystems, I just wanted to share the concept with you thinking maybe it would get some gears turning On another note there is a thread: here where a user put forth his space colonization otl that I could see also being very helpful in creating organic arc like shapes. Its a very similar process from what I did only its wrapped up nicely and fairly easy to understand if you take it node by node -N

You can also use node.parm("reload").pressButton() function Heres a shelf tool I made for myself one day to reload anything in my scene pointing to outside files: import hou, toolutils, os, sys def getNodes(type): return toolutils.findAllChildNodesOfType(hou.node("/"),type,dorecurse=True) def reloadFiles(): for node in getNodes("file"): ## Replace file with whatever operator type you want print ("Reloading: " + node.name() + " @ " + node.path() ) try: node.parm("reload").pressButton() ## replace reload with whatever the name of the button is print "Success!" except: print ("Failed To Reload: " + node.name() + " @ " + node.path()) continue reloadFiles() It may be a little overkill but I found it to work pretty well -Nathan

Now combine all the things!

haha goodtimes! If you wanted to break it out into 2 steps so you could do your fluid sim separately you were on the right track. Multi-solver, then wire object goes into the far left connector then the wire solver goes into the right connector followed by the sop solver. treat the right connector like a merge, with a left to right hierarchy where you are just combining the solvers . The main solver being the first one plugged into the right input, nothing needs to be plugged into either of those solvers. -N

several ways to do it. If you want a lot of control I would just use a wrangle node in a sop solver, point to your velocity sim say with input 1 so the code would be: vector vel = volumesamplev(@OpInput1,"vel",@P); @v += vel; then you can mult or do whatever do the velocity vector before its added to your wires velocity value. Could also combine popsolver and use advect by volume, but I figured this was simpler. -N

That means you hit your video card ram limit and if you want to simulate at that size as is, you need to disable opencl acceleration. -N

Thanks for having a look Tomas, I figured it was probably fixed in 15 -N

So before I submit this as a bug I just wanted to see if anyone else on here might have experienced the same issue. Houdini Version: 14.0.444 Step 1: Read in an animated cache (from wherever, doesn't matter) Step 2: use a split node to break something apart Step 3: plug the output of the righthand side of the split into a timeshift Step 4: Does the timeshift do anything? I have recreated this several times already. Every time I read in a cache (I've tried with many different types of caches), split something and try and use a timeshift on the right hand stream of the split, it does not work. Maybe this is a version specific thing, I haven't tested the latest release of 15 with this, I'm just curious if anyone else has come across this. Thanks -N

The Volume Visualization Sop is literally a node in sops called just that, it acts just like your guide settings within dops, it lets you visualize your fields and color them in the viewport and transfer back and forth from the pyro shader and the visualization node. That being said I would strongly recommend setting up fire and explosions with a physical blackbody model, rather than a hand picked color ramp, for several reasons. Consistency across shots and setups being number one, number two being that the physical model will have a much better color range when you get to your low and high values. But Yeah, as mentioned above that video is pyro 2 and if you are using houdini 15, you are using pyro 3 :-/ -N