Digital Art Zone

Carrara has indeed an excellent tree lab. But good trees are a poly nightmare. The important thing with Bryce trees is to set the leaf size small and ramp up the number. If trees are for the background, consider rendering some, including the mask, and make billboards of them. It is one of the strategies to bring memory usage down.

I would tend to agree with what Horo has said. Good trees require detail and that equals lots of polygons.

For this reason I’d say the approach one should take with trees depends on how close up one plans to view them. From a distance most trees, even Arbaro trees are great. Closer up, it get harder.

On the subject of Arbaro, did you know that this product in the store is nothing but Arbaro trees repackaged.

http://www.daz3d.com/shop/trees-for-the-forest/

Hands down the Tree Generator in Carrara is the most flexible tool available. The only issue is that in C7 and lower there was a bug preventing trees exported in obj format from displaying leaf uv properly. The bug is fixed in C8.

Bryce trees are good but not ideal for extreme close-up viewing.

Horo’s suggestion to use transmapping is a great idea. You can draw an image with a dozen small leaves on it and load it as a leaf source, getting a dozen leaves for the polygon expenditure of a single leaf. But transmapping is extremely slow to render, this gets especially true when there are compound transmaps. Basically, the more complex the svcene is with transmapped leaves the longer it takes. For this reason alone I tend to try to avoid transmapping. But sometimes blend transparency is the only way to go.

Dan, that render is fantastic!!!!!

I’ve been doing a bit of thinking lately on this and comparing what I see in rl with what I would have expected to achieve in artwork and have noticed something. One should really brake down distance of the viewed object in at least 6 levels; extreme distant, distant, mid, near, closeup, and extreme closeup. What is necessary for each of these levels varies vastly of course, but what is surprising to me on observation is how even ‘near’ doesn’t need as much detail as we might expect as our eye does not pick up the detail of the bark or leaves for instance. The trick it appears is that in rl there is a small amount of blurring going on even in near compared to closeup (30 feet vs 10 for example… varies from person to person a bit.) At 30 feet for instance, one could get away with a decent texture and a small amount of bump. At 10, one really needs a displacement map to look good.

3D software puts everything in sharp detail unless one uses DOF in their camera and, this isn’t exactly the same as DOF. Rather, this blurring runs in conjunction with DOF. Add to this motion blur which starts at small amounts even with small amounts of motion (like walking) and we see how this can add up. Our brains average this all out and we don’t notice it unless we consciously pay attention to it.

It would be interesting if we here went out over the next few days and did a conscious measurement of blur we notice under various circumstances and compared notes.

Final note on this, I think instancing and LOD are two areas that have a lot of promise in this regard. Having said that, extreme closeups will probably always be their own beast unless animating into it.

I mostly agree with Gedd… It amuses me when people put extreme detail (and therefore increase render time) in far away objects then blur them out with DoF.






I also laugh at myself when I do exactly the same.