Highlights from Vlado’s V-Ray 3.0 sneak peek presentation at EUE 2013.
Posted by Torgeir Holm on 1.7.2013 | Category: V-Ray news
[ UPDATED ] Vladimir “Vlado” Koylazov gave a presentation at the annual End User Event in Utrecht. He gave a sneak peek into some of the things we can expect to see in the next major version of V-Ray.
Interest was very high and the room was packed to capacity for the session, so he had to do it all over again an hour later.
Here are some of the improvements that we can expect (though no guarantees were made) in V-Ray 3.0:
Brute force rendering speed improvements
When V-Ray’s development was started more than ten years ago, brute force methods were generally too slow to have any practical use. Fast biased methods like the irradiance map and the light cache were developed, and as a result V-Ray was never particularly optimized for brute force path tracing. The biased methods also have side effects in the form of artifacts and flickering.
A lot has happened since the start, and with today’s hardware and advancements in algorithms brute force is now a more than viable option. As a result of this Chaos Group have spent time optimizing path tracing through internal code optimization, and also by adding support for the Embree raycaster (available when rendering on an Intel processor).
This is still an ongoing development, but examples shown already display significant speed improvements. The classroom scene below took 15m44s in V-Ray 2.4, and 5m55s in the 3.0 beta using Embree. That is 2.66 times faster brute force rendering!
Brute force rendering optimization (click to enlarge)
Max ray intensity
Another issue with brute force path tracing is the appearance of “fireflies”: single over bright pixels that will take excessive sampling to get rid of. These artifacts are very difficult to clean up and are usually due to hard to find paths caused by reflective or refractive caustics. V-Ray 3.0 adds a max ray intensity parameter that will reduce this problem dramatically. The option is similar to the “clamp output” option, but is done for each ray, rather than per image sample. It only applies to secondary rays, so information from primary rays is preserved. Although it is not physically correct and may remove some light from the scene, it may produce a clean image a lot faster.
Max ray intensity before and after (click to enlarge)
V-Ray 3.0 has a new image sampler called progressive rendering. This sampler works on your entire image, the same way RT does. Advantages over bucket sampling are that you always have an image to look at, and you avoid last bucket syndrome (those paying close attention at the presentation might have noticed another remedy for last bucket syndrome, one that was shown but never actually mentioned… We’ll leave that one for later... or comment if you know what I'm thinking of). Progressive rendering is controlled through a shading rate parameter that is a global quality control for GI, lighting, glossies etc. The progressive renderer fully supports distributed rendering.
Because V-Ray is increasingly being used for character work that involves lots of hair and fur, an effort has been made to optimize the hair rendering in 3.0. V-Ray 2.0 already added a specialized hair shader, as well as support for 3ds Max and Maya hair engines such as Hair & Fur, Ornatrix, HairFarm, Maya Hair/nHair, Shave & Haircut and Yeti.
In 3.0 the V-Ray hair shader has been optimized for semi-transparent hair, a requirement for soft and natural looking hair. Memory usage and hair raytracing have also received dramatic optimizations. The examples shown displayed a massive increase in speed, with render times going from 26,5 minutes down to 2,5(!) minutes for comparable results.
Hair rendering before and after optimization (click to enlarge)
Transferring V-Ray materials between different applications such as 3ds Max and Maya is currently a huge challenge. In a mixed application environment it would be very nice to be able to transfer materials as easily as we can transfer scenes and animation. Today’s method is very complicated and involves a cumbersome detour via the .vrscene format, and only supports transfers from Max to Maya, not the other way around.
At the same time V-Ray for Rhino and SketchUp have their Vismat materials. Chaos Group also have a huge library of Vismat materials on their site:
Vismats are XML based application-independent descriptions of V-Ray materials. They can be used to describe most materials normally needed. And though they do not support application specific features (in order to retain their application independence of course), a number of V-Ray specific procedural textures are provided.
Adding support for the Vismat format in V-Ray 3.0 for 3ds Max and Maya will allow for very easy material transfer. The Vismat material editor has a Python and QT based UI which is identical in all applications. A standalone material editor will also be available. A node based Vismat editor is also planned, but will not be in V-Ray 3.0.
Some indirect illumination effects are very hard to render with camera based raytracing. This mostly applies to caustics effects. Vertex merging is a very robust new GI algorithm that combines bidirectional path tracing and progressive photon mapping. It is a biased solution, but the bias is reduced with time. This method is also easier to use than photon mapping.
Vlado showed an example of a teapot encased in a glass box, and lit only by caustics. Without vertex merging the teapot rendered as a very dark object, not picking up the light passing through the glass. However, as soon as vertex merging was enabled the teapot rendered in a much more realistic way.
Vertex merging (click to enlarge)
V-Ray RT GPU
V-Ray RT GPU now supports render elements, allowing it to make the transition from preview tool to a production rendering tool in a compositing based pipeline. V-Ray RT GPU supports most of the render elements (though not all). One improvement over the regular V-Ray renderer’s elements is that the GPU version supports full light select elements, including GI, light, reflections etc. per light selection.
Another very interesting V-Ray RT development is that V-Ray RT can now be used as a Maya viewport renderer. This seems very useful! Unfortunately the same is currently not possible in 3ds Max due to restrictions in that application, so 3ds Max users should start nagging Autodesk to change this!
The viewport mode in Maya supports all V-Ray RT features, including CPU and GPU rendering, as well as distributed rendering. The render resolution can also be adjusted for added speed (i.e. render at half resolution).
See a quick demo video here (this video probably does not contain the most recent developments, as it is a few months old):
Automatic asset transfer for DR – No more missing maps EVER!