Well...my XSI experience CD has arrived and I thought i'd share some of my thoughts on the entire set.
One thing that hits you about XSI when you first load it up is the complete neutrality of the interface.
Theres' your standard animation tack along the bottom, object modules along the left side and transformation and property controls along the right side.
The viewports are completely customisable, you can drag them out any which way. By clicking on the name (perspective, front, left etc.) you can access other fuctions like synoptic view, the rendertree (XSI's material editor), the FX tree (an inbuilt compositor new in 2.0) and the Animation Mixer (this thing is just MAGIC).
The Experience set is a limited demo - its basically what you'd get if you shelled out 8 grand for the commercial version. The only two provisos are, whenever you render its' watermarked and you can't save. at all. which sucks, but is kind of understandable. The second CD in the set contains a huge, huge HTML page with about 70 tutorials in WMA format (i think). The tutorials show you how to use some of the new functions in XSI 2.0 as well as give you a very comprehensive idea of what the whole program is actually about.
Moving things around in XSI is pretty easy...however once thing is you HAVE TO GET USED TO USING KEYBOARD SHORTCUTS. unlike 3ds-MAX where you can use shortcuts or you can use the buttons that so unergonomically cram the screen space, to move the camera for instance you have to hold down "S" plus one of three mouse buttons. Its a little odd, but after you get used to it, it works so well you wont want to go back to anything else. Moving, Rotation and Scaling is handled by three panels on the right hand side of the screen. You can move it in the viewport or use the type-ins, both feel responsive and with a comprehensive set of grid, vertex and line snapping tools you feel like you can do anything with it.
Anoter dissimilarity to MAX is the way it handles sub-object modes (you know...where you collapse your object and start editing it on the vertex/edge/face/poly/element level). In XSI you dont have to convert it to anything, as each little change you make works like a new modifier and is saved in the "construction history". At any point you can go back and delete some screwy thing you did that messed it all up, this does mean that if you've done
of work on a model it can get a little slow if you dont delete the history or "freeze" the object as an XSI-er would tell you.
XSI supports great polygon modelling tools. By pressing spacebar you're in object mode...hit "U" once, and you're in polygon mode. Its worth mentioning the way the selection works. In MAX...I would have to hold control down and click each bloody polygon to select what I wanted. In XSI, i just hold down the mouse button and paint my selection on. It sounds silly, but it really is SO much faster its' unbeleivable. XSI will not make a mistake and select backfaces either. Essentially the selection system is way more elegant and nowhere near as clunky as MAX.
Most modelling options are supported, extrude operations open up a whole new dialog window allowing you to set the amount of subdivisions on the extrusion (more useful than it sounds!), and in addition there are a set of amazing scaling and rotation tools. There are very useful polygon bridging tools as well an "extrude along path", which lets you choose a spline as a loft to extrude along.
When you've finished modelling, just hit the plus key on the numeric keypad and voila!. XSI's vaunted SubDees kick in and you have a nice, clean subdivided surface. None of the rubbish about triangles and octagons causing smoothing artefacts here, everything is as smooth and gorgeous as if the model was made out of equal quads. Which in most cases it is anyway, since XSI makes it hard for you to mess a model up the way you can in MAX.
Next up is the materials editor and rendertree. The RenderTree is a node/flow based diagram where you connect up nodes to "ports", for instance;
Your shader (the peice of code that tells the renderer what colour, specularity and surface characteristics the geometry has), blinn in this case, has a number of slots. Diffuse tells it what colour the object it, Specular how shiny, bump simulates an uneven surface and so on. In MAX you'd click on the empty slot and a few dialogs would come up. In XSI you select whatever texture you'd like from a dropdown menu, click on the little red dot and connect it up to whichever port you'd desire.
Unlike MAX, XSI sometimes requires "conversion nodes" for things like 3D textures, projections and Raytracers because the nodes that generate those effects are Scalars, not textures. Its just a simple case of going to the conversion drop down and making a conversion node and plugging it in. XSI also supports directX 8.1 shaders for realtime work. Since XSI is openGL compliant ONLY, the DX8 shaders are working in emulation mode and therefore you can see them regardless of what openGL compliant graphics card you happen to be running.
The rendertree also controls things like volumes, which again are mental ray shaders. No more messing around with plugins like afterburn etc, just assign a volume shader to a particle and smile. Mental Ray chomps through volumes like no other and the quality of the results are gorgeous.
The great power of the rendertree is your object can have a different shader for different functions. the surface could be blinn, the shadows could be generated by something else and the photon/global illumination could be generated by something different. This gives you a great level of control over what visual results you can acheive, no longer being limited to what the renderer and application approximates your settings to. It is also worth mentioning that all the elements of the XSI render tree are Mental Ray shaders and phenomenon, and apparently its' quite easy with the SDK to add your own in.
Rendering in XSI is sublime. Instead of in the materials editor where you have "preview spheres", XSI takes renderer integration a few parsecs ahead of the pack. Press and hold down "Q", drag a window out over your object in perspective. Something incredible happens.....mental ray kicks in and renders the region as you'd expect to see it. The region controls and the "traditional" render controls are completely seperate (although there's a copy settings script) for fast previews. Anything Mental Ray and XSI can do in the final render is available in the RenderRegion. This is an amazingly powerful tool....I miss it already when I have to go back to MAX (no, Mayas IPR and MAX's ActiveShade dont hold a candle to this!!)
Animation in XSI..well...i've barely even scratched the surface here but suffice to say its apparently one of the best bits. The mixer works in a more "organic" fashion to MAX's keyframe-based setup. You essentially assign a pose, for instance a door open and a door closed as a "clip". These you can put into the Animation Mixer and generate a transition between them. This isnt really that amazing. However, clips themselves can be smaller animations. Supposing I had a guy running, and a guy sneezing and generated a transition. Works flawlessly. As I said above, its a very organic way of doing things and saves huge amounts of time. Needless to say, other animation controls have useful little tweaks. SUppose i wanted a spaceship to follow a path towards the camera? XSI gives me an option called "use tangency" where the ship will align itself to the tangent of the spline. Brilliant! no more messing around with keyframes to make sure its not pointing backwards to the direction it should be facing! Inverse and Forward kinematics can be mixed with no problem, and Softimage has long been the best package around for character animation.
Finally, simulation.
This is normally plugin territory for just about everything, and unfortunately XSI has yet to be given some rigid body dynamics controls.
However, it comes with ergonomic and fast soft body and cloth simulation modules which produce VERY realistic effects and as you'd expect, support a level of interoperability that puts other packages to shame.
Also new in 2.0 is the hair simulation system. I havent used this yet, but it would have to be totally useless to be worse than Shag:hair, and after watching the tutorial videos I know its way better. I beleive the system was based of Joe Alters' "Shave and a Haircut" system, except the hair is generated as mental ray geometry shaders and as such, you can have massive polygon counts with reasonable rendertimes. Instead of just guide hairs, its possible to generate openGL interpolated hairs which give you a great idea of what the final thing will look like WITHOUT having to waste time rendering it all.
All in all, XSI is working towards becoming a true all-in-one solution. And the elegance of the interface leaves other packages in the dust.
Hope you found this review relatively useful and informative
wEvil is an Animation Student at a British university somewhere
He has almost five years of experience using 3D studio MAX since version 2 and is currently working on a movie project in addition to his studies.
