Green Glass Dodecahedrons

I started working on making a frame from my Dodecahedron Rain into a nice still image with glass dodecahedrons and a floor texture two days ago. First, I had to work on putting a material on the dodecahedrons, and this was new to me, since I had used a dynamic simulation to clone the objects, instead of a normal clone operator. I will not go into deep details, but it took me a little while to get the material to work because of the odd configuration. After I learned how to get a material on a dynamically cloned object, I had to find the material I wanted, and then configure and fine-tune it to suit my needs. Once I had done this, I found a frame that looked like it would make a nice still image. I then set up the lighting, using one point light to generate photons for caustics, and an area light for shadows and normal lighting. With the lighting all ready to go, I went ahead and generated a photon map. Luckily, I was familiar with the controls and this took little time to configure. After I let to photon map render, I double-checked all the settings, saw I was ready for a final render, and started rendering.

As of this writing 53% of the image has been rendered and the rest is still being rendered. When I set up the scene, I knew it would take a long time to render, since I allowed a maximum of 10 reflections and 10 refractions for the glass material, and there are a few groups of multiple glass objects in the scene. Tracing a light ray through multiple reflections and refraction can be computer intensive. However, I never expected a time as long as this; it has been rendering for 17 hours as of this writing. I also think that the second half of this image will take longer to make, because there are more glass dodecahedrons in it. Therefore, it could be another 24 hours at least before I get a finished image. The first half looks good, so I am assuming the rest will.

UPDATE: It has now been 33 hours and 30 minutes since I started rendering this image, and it is 73% done.

UPDATE 2: 36 hours after rendering start, the image is 85% complete, and I think there are no more densely populated parts of the image, so it should be done by this afternoon; in roughly 6-10 hours.

I hope that nothing happens while the image is rendering, as it is going only to my RAM, and not onto my hard drive. This might not have been the best decision, but I wanted to be able to look at the image while it was being rendered, and as far as I know, I can only do this without saving it to my hard drive as it is rendered. If I were saving it to my hard drive as it rendered, and something crashed, I would at least have half an image, and I could just render the rest of it, instead of all of it again. I am hoping that sometime I will find out how to make mantra, the rendering program, output to two places, such as a live view, and a file on the hard drive.

On a side note, my laptop processor has been overclocked by 21% this whole time, so I guess this is a nice stress test too, and so far, my processor is passing. Although the temperature is getting high, with one core at 78°-80°C (172°-176°F) and the second core at 74°-76°C (166°F-169°F), assuming my temperature readings are correct, which I believe they are.

FINAL UPDATE: After a hot (for my processor) 38 hours, the image is finally done. I will post it soon. I was surprised at how fast my processor temperature dropped; it went down 20°-25°C in about 3 minutes.

To see the final render, go to http://flickr.com/photos/heydabop/2947837408/.

Houdini’s RBD Simulation

After doing many complicated (to me) projects, I chose to do something simple that I assumed would not take long to set up, a simple Rigid Body Dynamics simulation. It consists solely of dodecahedrons being generated at random spots and falling to the ground. It took only fifteen minutes to calculate the 600 frames of the animation. It could have done it even faster, but it saved every frame to disk, which apparently takes a lot of time to do, because the first 31 frames took a minute, yet I can make it calculate the first frames in real time at over 24 FPS if I don't save to disk. I rendered it at a low quality, since it is not anything special in the first place.

Final Video: http://vimeo.com/1942465

I might do a single frame render later on, where all of the dodecahedrons are glass, with a floor texture, better shadows, and a higher resolution.

Collisions 2

After posting my problem that I explained in my last entry in the Side Effects forums, I received an answer from Allegro, who I believe is a Side Effects employee. From what I have seen of Allegro's posts, he seems to know what he's talking about. So I read through what he had posted, understanding most of it. I could not look at the file attached to his post, as I was short on time.

I returned later and downloaded his file. I opened it to find a high quality collision mesh of my geometry. Pleased with what Allegro had made I went on to fine tune my fluid simulation. After computing the simulation, I then foolishly rendered it without actually watching it all of it. I had to do some test renders of the first frame to get lighting down, and get a good render time to quality ratio. After I set up the lighting and render quality, I rendered all of the simulation. I kept checking on my computer every few minutes to see how far it was, and see if everything looked good. It took about two hours to render, and once it was done, I played the finished sequence. I was not happy with what I saw. You could only see the water drop in two frames of the 24 FPS sequence, and it settled in its container in about ten frames. Frustrated with myself for not checking my simulation before I rendered it, I started to think about how to slow down the drop. However, I quickly decided to work on it later. So far, all I have thought of doing is increasing the scene's scale. I will see how that works in a day or two.

Thanks for fixing my collision mesh, Allegro.

Collision Thread: http://www.sidefx.com/index.php?option=com_forum&Itemid=172&page=viewtopic&t=13588

Collisions

I have been trying to make a scene in Houdini where a drop of water falls about three meters, and then splashes into an oddly shaped container. It has not been going as well as I would have hoped. First, I had to make it so the particles were not being formed uniformly, because then I would get a uniform splash and it would look fake. Looking though controls on the emitter node for a little while solved that problem when I found jitter. I then added drag to the particles, and I was inexperienced with drag, so that took a little learning too.

Now, most of these problems I have just listed were not too big, and were easy to solve. However, I have one big problem that has been hounding me throughout this whole project, and I have yet found a way to solve it. The container that the drop of water is falling into is not a simple or normally shaped container. It looks somewhat like a modern squashed glass vase. This is hard to represent well in a collision mesh, and still being relatively new to Houdini, I am having trouble doing so. It would be easier if I did not have to represent interior features, but I do, or else the container is not container, and is then worthless. I have been using the trial and error method, but this is not the best method, since each collision geometry takes anywhere from 30 minutes to 3 hours to generate. So, I have posted my problem on the Side Effects Houdini Forums, which is full of great, helpful people, and hoping someone there knows a way to get an accurate collision mesh for my geometry.

Container: http://img56.imageshack.us/img56/5651/containeret0.png

Resulting Collision Geometry: http://img56.imageshack.us/img56/9083/collisiongeometrydr3.png

Global Illumination

Over the past three days, I have been doing a little work on Global Illumination in Houdini. I made a scene with bright colors to emphasize the diffuse reflections. I had to do a lot of tweaking to get a result that was similar to what I wanted, still not exactly what I want though. So here they are.

 

This is my first try. The main problems with it are the red boxes reflection to the right is to bright and has weird edges, and the texture on the tube is weird and has a seam on the right side of the image. Other than that, it came out decent.

http://flickr.com/photos/heydabop/2923490110/

 

For my second render, I changed the lighting upped the photon count to 1,000,000. I also added in caustic calculation to help keep the Global Illumination from being over lit, as with the reflection in the last image. The caustics are not rendered in the final image. In lighting, I went from one spot light, to four spot lights, creating a square behind the camera.

http://flickr.com/photos/heydabop/2923495580/

 

I decided to change the tube's texture in the third render, since I am too inexperienced to get the UV projection I wanted, I just went with a homogeneous texture. Same lighting and photon count as in the last image. I am going to render this again, but get rid of the bronze tint from the tube, as it does not suit my tastes. One thing I found interesting in this was the purple reflection in the blue box. The only thing I can guess is that the red box is reflected on the back of the tube, which reflects on the blue box.

http://flickr.com/photos/heydabop/2922647089/

Same Torus, New Light

I decided I go ahead and try to re make the Glass Torus scene in Houdini. I chose to leave out a floor texture because I couldn't find one that I thought looked good, and it greatly increased render time and photon map generation time. However, I changed the lighting in the scene to get better looking shadows. I used a spot light for photon map generation, and I used a grid area light for shadows. It took about thirty minutes to generate the photon map for caustics, and then 11 hours to render the image. I like it how it looks; I just hope other people do. It's simple, yet still looks good. Any intelligent suggestions or criticisms are welcome.

http://flickr.com/photos/heydabop/2912048362/

 

Now that I've finished this I'm going to try a scene with prominent Global Illumination.

Recap; Side Effects Houdini Apprentice HD

As you can see, what I predicted when I started this blog happened to some extent. I got lazy and did not feel like writing. So here is a quick recap of what happened over the past two months.

Hurricane Gustav hit me hard and parts of it were scary. Streets near my house flooded, and we were without power for 5 days, with is not much compared to many other people hit by Gustav. We had a lot of tree debris to clean up. It took us about 2 weeks to get most of it out of the way.

A week and a half after Gustav hit I went to Disney World. We did not go because of Gustav; we had been planning this trip for almost a year. We spent a week at Disney World, in a Disney resort on Disney property. It was one of the best weeks of my life. I spent 3 days in EPCOT, and one day in every other park except for the water parks, which I never visited. The last time I had been was in 2000, so I remembered a little bit of Disney, but not much. They have changed a lot of stuff too, so most of it was a new experience to me.

Besides those two major events, life has been moving at a normal pace for me.

 

Now, on to Side Effects Houdini Apprentice HD. Houdini is a 3D animation and modeling program made by Side Effects Software Inc. I stumbled upon it after doing countless searches for software allowing me to do computational fluid dynamics. When I first saw it, I just passed over it, remembering using it a few years ago, and not liking it. This was partly due to my inexperience. After running out of solutions, I finally went back to look at it, and watched some demo reels on it. Amazed by what I saw I decided I had to try it out. I downloaded a free version of it called Houdini Apprentice. It limits your rendering resolution, and puts watermarks on everything you render. After messing with it for a few days and doing some tutorials, I bought Houdini Apprentice HD, which takes off the resolution cap on still images, raises the resolution cap to 1920x1080 on videos, and gets rid of watermarks. I have now owned it for about 2 weeks.

The more I learn about it, the more I realize there is more to learn. It is full of features for almost anything I could need. It does very well with dynamic simulations, but those take a lot of processing power, so I have not done many with my 1.8 GHz AMD Turion 64 X2 over clocked to 2.2 GHz. I am still learning so much about it, and still being impressed by what I see. So far, I have only made two scenes worthy of release into the public domain, portrayed in the thumbnails below. I hope you liked them.

 

The highlight of this is the use of caustics. Rendering caustics in Houdini proved to be trickier than I had expected. Many times when I rendered it using Physically Based Rendering, not only was there noise created by PBR, but random bright white specks and blocks of pixels would appear. This was not acceptable, so I went posted a topic on the Side Effects forums, and got no response. The forums are very helpful, and have assisted me many times before, but I guess no one knew what was wrong here. So I went looking around for people with similar problems, and while doing this I found a way to render caustics without using PBR. I figured this would at least get rid of noise, so I went ahead and did so. I then rendered the scene again using Micropolygon Rendering, and got what you see here. I plan to render this again, improving caustic count and quality, and make giving the floor texture. However, this will take much longer to render, and I will probably have to leave it over night. Hopefully, the result will be worth it.

 

This is the first scene I made for rendering a still image. Making this was a multiple stage process. First, I had to model the Möbius Strip, which included learning how to model objects. After modeling it, I rendered the first image, with the textures and no wireframe, which is what you see on the right of the image. Then I had to go and learn how to do a half-wireframe render. This took some searching on the forums, and combining parts of what I found. Once I was successful in rendering the wire image, I went into Paint.NET and combined it with the first render, and used an alpha gradient to make what you see now.

For more Houdini renders, go to http://flickr.com/photos/heydabop/sets/72157607632268880/.