johnLIC

My bad. I had @pscale -=.01*rand(@ptnum) in a point wrangle node. I guess the sand upres is smart enough to keep track of particles, even though the @ptnum is changing. Sorry for the distraction.

I made a grain solver simulation, then deleted many stray points from the cached .bgeo. When I do an "upres sand" the points that get generated are different on each frame, so the part of the sim that is supposed to be standing still is sizzling. I think it has something to do with point id's changing every frame. I found a solution from Jeff Wagner that is a script he wrote in 2009. I haven't tried it yet, not sure if it's what I need. Is there a better way?

I'm caching a Pyro sim out as VDB. I want to preview my work, but when I read in the VDB, using a fileCache node, it is a mostly white blob. How can I bring in VDBs so that they display correctly for flipbooking? (They render fine) Or, is there a conversion I need to do? I'm optimistic there's a solution I'm just not familiar with.... SOLVED: volume visualization node set to read density.

I want to use the u or v tangents of a NURBS surface in an attribute wrangle node. What are the names for the tangents? Well, I found the polyFrame node. It doesn't play nice with NURBses, not surprisingly, but if I switch to poly geo I get some tangents. I still want to know how to access the tangents of a NURBS surface, but at least I'm back in business now...

Thanks! My Nvidia GTX 970 has 4GB of GPU RAM. In that demo file I can crank the Uniform Divisions on the smoke object to 210 before the smoke solver stops working. That gets me 9,600,000 voxels. I'm using about 27 million voxels before I get a decent render. Maybe I can split the space up into smaller chunks....

Does anyone have general advice for a long shot with a huge zoom out/pull back? I'm trying to create some lingering dust. I can simulate at high enough resolution to get a decent rendered look when the camera is pulled back and we see the whole sim, but I've got almost 200 frames at the beginning of this shot where the camera is zoomed in and looking at a small section of the sim. That part doesn't look like much right now. Is there any alternative to simming a whole dining room with 3 mm voxels? Any advice on slow moving, lingering smoke?

Or model the deforming tornado whirlpool shape ( maybe in NURBS so the tangents can drive the vel?), and sim on top of it?

Well, openGL is not as pretty as the renderview images, so I kept looking, and found this awesome script: https://github.com/ArthurYidi/Houdini-Quick-IPR-Preview Here are complete instructions, not long, but maybe not intuitive for everyone... I put the two .py files in this directory, then restarted Houdini: /opt/hfs16.0/houdini/python2.7libs I imported the new modules in a Python shell in Houdini: import tiff import renderIPR This script needs to know the name of your Render View. In the Python Shell I did this: p = hou.ui.paneTabs() print p and then inspected the contents of p manually (well, with my eyes, so... visually?). In my case the Render View pane was p[3], like this: <hou.IPRViewer panetab3>. Knowing the name of the Render View pane, I created an IPR object, as defined in the file renderIPR.py. I named it J: J = renderIPR.IPR(p[3]) I set the frame range for a quick test: J.setFrameRange(0,10,1) The first time I tried: J.render() it failed saying invalid plane. It was looking for the color plane in the Render View, but I hadn't rendered anything yet. I rendered one frame with the "render" button in the UI, then tried: J.render() again, and it worked! It made a render directory and dumped 10 frames in there. I hope this is helpful to anyone wanting to save out a render view frame sequence.

Thanks! This is great!

Yeah, I guess you're right. I can't even get a crappy frame out in under 15 minutes. Does anyone have a script that will render a frame in the renderview, write it out to disk, and then advance to the next frame? I don't have 300 CPU hours available, and I really could use the renderview frames, at least for now. I'm looking at a list of Python commands now....

I have a cached pyro sim I'm trying to render. When I render a test frame in the render view panel it takes about a minute. That's using the mantra node, without preview on, using micropolygons, with 2 pixel samples, noise level = 5, and other stuff turned down for fast rendering. I clicked "render to disk in background" and it's been running for 32 minutes, and doesn't seem like it's going to stop any time soon. This is a fresh mantra node. The previous one was just as bad. I've got two 2048x2048 depth maps for shadows. Could that be the problem? I'm stumped, but I've got 600+ frames to render, so any help appreciated.

The chop needs to be tuned according to the scale ("ampscale"). At lower scales it takes more chop to get peaks.

I'm having a go at it, but downsampling 3 or 4 results in that quilted, square patches look. I'm just having a hard time imagining how it's going to work conceptually. If I could gamma the map so the high values are higher, that might make the waves crisper, but I'm not sure how to do that between the ocean spectrum and the ocean evaluate. I'd like to do it to the output .exr map in Nuke, but my non-commercial license only outputs 1920x1080...

I'm trying to make an ocean surface about 10 meters across, with a quiet feel. I'm having trouble making waves that have peaks without a lot of higher frequency displacement. How can I get waves that are peaked but not lumpy? I need to export displacement maps to be rendered outside of Houdini, so I don't have the advantage of the Houdini ocean shader. I think I've got enough resolution in my maps (3000 x 3000). When I try to filter out the lumps I lose the peaks and it looks like a rubber sheet.

http://www.johnpattersononline.com/