Softimage’s next-generation 3D package,
XSI, has had a difficult birth in the highly charged high-end 3D marketplace. Version 1.0 was late and missed essential features like polygon modelling. In fact, it was missing so much that Avid was forced to give away a full version of Softimage|3D to paper over the cracks. Version 1.5 added the polygon modelling and smoothed over some of the workflow problems, and now we have version 2.0 (at least in early-release form), which is looking very slick indeed.
The Softimage|DS compositing application has traditionally been partnered with XSI, but now version 2.0 of XSI has its own compositing application built directly into the interface. The new compositor is based on the Avid Illusion compositing engine, and works on a node-based metaphor, similar to XSI’s own RenderTree method. The compositor ships with 65 film-quality effects that can be applied at 8- or 16-bit or even at float precision. This addition will be a real boon to animators, since it avoids the hassle of switching applications – everything can be done from the same seat, and previously tedious-but-necessary operations like multi-pass renders can be managed from within the application easily
A front end that suits
The interface remains extremely customizable, and different layouts can be saved and recalled for different stages of the production pipeline. Also, Proxy pages
are user-defined parameter sheets that can be tailored to contain sliders relevant to parameters of the job in hand. This saves the trouble of having to dig down through an object’s explorer in order to edit the desired values. The new Synoptic view also allows users and technical directors to create customizable, HTML-based interfaces with embedded links. A character’s representation (which is just a bitmap) in the Synoptic view can have scripts that are linked to representations of the result of the script (bend elbow, and lift knee, for example). Choosing the appropriate pose from the Synoptic view executes a script that triggers the relevant response in the model in the workspace.
The Proportional Guide is another new feature in version 2.0 that will ease animators’ lives by making character skeleton setup much easier. Softimage’s products have always included ready-made character skeletons of varying complexity, but the Guide is a high-quality skeleton rig – even including properly-defined lumbar movement – that animators can adapt and mold to their own ends. Once the rig is positioned inside the character’s mesh, the click of a button turns it into a fully operational, bound skeleton with all effectors in place. At present it’s only available for bipedal characters, but a quadruped is coming.
The other great workflow enhancer in XSI is NetView. This is essentially Internet Explorer running in a window inside XSI. It gives easy access to model libraries, walk cycles, textures and the like from Softimage’s own Web site. Version 2.0 also has a
plug-in for searching the excellent TurboSquid 3D site, currently the best 3D repository around.
Subdivide and conquer
XSI 2.0’s polygon modelling makes full use of Subdivision Surfaces. This is a method of deriving smooth, organic objects from blocky, polygonal cages. It’s very much de rigueur at the moment, and XSI’s implementation is about the best there is. Two methods of subdivision algorithm are provided, Catmull-Clark and Doo-Sabin, and different levels of subdivision can be set locally on an object’s surface. This highlights one of the great features of Subdivision in XSI: you’re not limited to triangular or quad patches in a model, which increases creativity and productivity considerably.
The modelling workflow is excellent, and puts
the rather lacklustre implementation of Subdivision surfaces in Maya to shame. For example, selections made at different pick levels (face, edge, vertex) remain live: change the pick level, and the last selection made is still highlighted for you. Also, symmetry modelling is back, where changes made on one side of a model are immediately reflected over to the other side.
Weight painting is a very accessible way of modulating the effects of bones on a character’s mesh and version 2.0 has added a dedicated Weight Painting toolbar. The effects of a Weight map are now shown directly in the interface, with the map colour corresponding to the bone’s default colour. The sphere of influence for a particular bone can be adjusted by simply painting on (or erasing) its influence colour in the workspace. The model mesh will update in real time as the influences on it vary.
Also getting its own share of the workspace in version 2.0 is simulation. It’s tightly integrated into the application, or at least it is in the Advanced version. The Fur and Hair simulation has received a lot of attention, and Softimage’s engineers have found a way of relating hair characteristics to texture maps: The controlling effect of a texture map can be applied along the length of an individual strand so, for example, a character’s hair can be straight near the scalp, and get progressively frizzier along its length.
XSI’s texture placement tools have also had an overhaul for version 2.0. Mapping textures around complex geometry is always a chore, but XSI has tools to make the job relatively easy. Sub-projections allow areas of a character’s mesh to be detached, dragged and scaled to fit over the relevant part of a texture map sheet, as before. However, a new ‘Heal’ tool has been added to eliminate the dreaded seam problem. This intelligently matches up vertices that were separated in the original phases of texture-map assignment, automatically eliminating seams.
Version 2.0 adds the ability to assign multiple UV projections on the same piece of geometry, allowing coincident texture layers to have differing mapping directions. Another texture-based enhancement is the ability to hook directly into the pixel shaders on some graphics cards, which gives full hardware rendering of textures and transparency effects.
XSI is still shipping in two versions, Essentials and Advanced. Advanced adds support for particles, dynamics simulation, fur and hair rendering, and a two-CPU licence for the Mental Ray rendering engine.
Version 2.0 is also the first version to run under Linux – and the first to truly fulfil what we were hoping for from the long-awaited Sumatra.