Chandelier help

[jophus14]

I am looking for suggestions on how to create the chandelier in the attached image. My chandelier will be 2' X 15' in size. It's consist of small metallic beads that form a strand and hang from the fixture. The only way I can think of constructing this fixture is to create a sphere, create strands from that sphere and then copy the strands all the way around the fixture. The problem is that I will have 10 thousand spheres per fixture and I will have 12 fixtures in my image. I need some help and hopefully someone on here can help me out. Thanks.

[percydaman]

how about ripping that image up and mapping that image to some simple geometry?

[joconnell]

Indeed - if it's that scale why not just use a vray 2 sided material on two cylinders - as if you were making a lamps that had a material shade on it. If you're making it the other way then make some cylinders and birth a sphere particle at each vertex with pflow - snapshot to a mesh.

[jophus14]

It's actually going to be a 2' X 15' rectangular shape light fixture.

Percy: I am going to try to map the image after some cropping and copying and see how it looks.

joconnell: I have a few images from the same website of this fixture with lights turned on and off and I will try using a 2sided material on it to see the outcome. I'm using VIZ2007 and I don't think I have the option of using Pflow. Wouldn't having a sphere particle with all of those cylinders still produce a huge file?

Thanks for the quick responses too.

[joconnell]

It will indeed make a huge file but vray handles the geometry really well - I was using vray fur for the needles on a christmas tree and although it made viewport life a bit slow, it rendered far quicker to replicate it with pflow and collapse to a mesh - I used the spacing tool with a sphere to make beaded mantle piece ornaments too. You could use a scatter to do the exact same thing since you don't have pflow. If you collapse to a single object you can turn on display as box to speed up your views.

[percydaman]

the great thing about that photo you posted is that the chandelier is straight on to the camera. Perfect for taking it into photoshop and cutting it out of the photo. Slap it onto the simple cylinder geometry and set it 100% self illum. Maybe throw the diffuse image into the self illum slot and check the color box. Dial in proper self illum from there. I think it would work smashingly.

[jophus14]

Here's the other image that I have:

[joconnell]

In this case it depends on how much you'll see the individual balls - you might be loading in a tonne of detail for something you won't get a lot of value from. But you'll probably get the look you want quickly. Faking it as a cylinder with the two sided map will be less polys but might take more tweaking on the material. It'll render far quicker when done though and the knowledge will stand to you.

Alternatively as percy suggested, comp it in if you have something high res enough - instant results for free.

[joconnell]

Actualyl nikki has a similar look to these lamps and he used a vray 2 sided material - http://www.chaosgroup.com/forum/phpB...highlight=lamp

[jophus14]

The chandelier turned out pretty good so far. I will post an image when I complete the entire scene. I have one more question regarding Vray Proxy that I would appreciate if you guys can help me out with. The scene will contain tables, chairs and table settings (silverware, plates, centerpiece, etc). What would be the best way to set up a VrayProxy for the tables and all that goes with it. Since there will be a number of items on the tables, should I set up a Multi-Sub material for all of the items or should I have all items seperate and then create a Proxy for each item with a material for that particular item? I'm afraid of having a Proxy with too many materials attached to it. Thanks

[joconnell]

Does it definitely need proxies? I'd try to render it without them first, they add a lot more time to your renders.

[jophus14]

I had just 39 tables with 8 chairs per table in my scene and I ran out of memory (2gb). I know that if I add all of the items to the top of the tables I will be in trouble. I created one table with 8 chairs and copied them as an instance and the computer kept on crashing. My scene isn't complicated at all, but for some reason I am having a lot of problems with it. I did a test render overnight only to come in the next morning (16 hours later) to find out that it estimated that it needed 36 hours left to render out. I had to drastically change my setting and I let it render out and this time it took 12 hours to finish. One wall has a reflective material and some sections of the ceiling have a reflective material, other then that it mostly tables, carpet and drapery. Here's some attachments:

http://i146.photobucket.com/albums/r...14Ballroom.jpg
http://i146.photobucket.com/albums/r...4/Settings.jpg
http://i146.photobucket.com/albums/r.../Settings1.jpg
http://i146.photobucket.com/albums/r.../Settings2.jpg
Ceiling Material: http://i146.photobucket.com/albums/r...ngMaterial.jpg
Wall Material: http://i146.photobucket.com/albums/r...llMaterial.jpg

[percydaman]

adaptive subdivision takes up alot of memory compared to adaptive qmc. Try precalculating your irrad map then using adaptive qmc for your AA.

Also try:

Lower your bucket size

Render out to vrayimage format

Turn on dynamic memory in system rollout instead of static.

Use proxys

Render using backburner (will use less memory compared to rendering locally)

BTW .85, .003 is pretty much overkill for qmc settings. It was before vlado readjusted qmc and it is even more now. I frequently use .94, .05-.15

Also your subdivisions on your reflective glossiness is overkill at 50. That was only really done when you used interpolation. I never take them above 20, usually 16 with decent qmc settings.

[jophus14]

Thanks for the info Percy. I made the changes and I will render the image overnight again and compare render times. I am unable to use backburner because I am unable to distributed render since VIZ and Vray are only on 1 machine. I might be wrong but I believe using backburner is only for distributed rendering, right? I lowered my bucket size to 32,32 and the other changes that you suggested. The reason why I chose Adaptive Subdivision as my AA is because of the amount of solid, non-reflective materials that I have in my scene. I was told to use Adaptive QMC when the scene contained mostly reflective materials. What do you think as far as Vray Proxies go with the table and table settings? Should I try to create a multi-sub object and Proxy for the table, chairs and settings or should the table, chairs and table settings objects be a proxy indiviually? Thanks so much for the help.

By the way.........What do you think of the chandeliers?

[percydaman]

chandeliers look pretty good. A little drab though.

There's another reason for adaptive qmc, and thats the reduced memory overhead it requires. The more glossies you have the more you want to look at using adaptive qmc, but also there's the memory advantage. Mostly likely itll render slower then adaptive subdivision if you dont have alot of glossies, but rendering slower is better then not rendering at all. Using proxies and switching to dynamic geometry is going to slow things down as well, but once again, its a lesser of two evils.

With regards to proxy, the most efficient memory wise is to have the least number of unique geometry objects. So would be to make one chair a proxy and instance it around one table made proxy. You then instance those 'proxy' sets around the room instancing again. Same goes for the place settings. Create a proxy from each object in the place settings and then instance them around the table. So in the end, you have just ONE of each object thats unique, and ALL the copies around the room are just instanced from the original object.

With regards to backburner, you can still use it with just one machine. Just start up the server and manager on your machine, send off the render to backburner, choose your machine and set job to start as suspended. Then close out max and activate your job using the queue monitor. You've now saved several hundred megs of memory cause you didn't have to load all the UI and stuff max needs to operate in the scene.

Hope that helps.