Techniques are like asshole and opinions, everyone's got one and everyone thinks everyone else's stinks.
I think it's time we update each other with the techniques we're using to cope with the current gen of graphics.

Here's mine. Assuming I'm modeling from scratch:

-Use splines and surfaces to build up dense geometry fast, while retaining control over the contour of the surface by editing the spline cage. In Maya, you can tweak the spline cage and have local control of the polygon surface at the same time.

-Draw Split Tool, plugin to draw topology directly on the mesh. Works really well with a Wacom.

-Sculpt Polygons Tool, to make sweeping changes to a mesh's contour without going in vert by vert. Also works great with a Wacom.

-Lattice, to quickly change the base volume of a mesh.

There is still room for extrude by edge and box modeling, but these are techniques I like to use later on when the mesh starts to get dense.

By the way I use Maya 4.5, still holding up well after all these years. Hasn't been updated much either, but that's besides the point [img]/images/graemlins/laugh.gif[/img]. I can't wait to hear from the Silo 1.3 guys in particular. Bet they have a lot to crow about.

I'd box-model in the rough masses, using lattices to vary proportions if necessary, then do detail passes over the whole model with polygon editing tools (Cut, Connect, Remove, Collapse etc) until it was suitably detailed.
Then again, that's how I model pretty much everything.

for complex organic things , i tend to start with a profile and then extrude edges to get the general shape, then cut connect etc where i see fit ... for things that i think are more regular in shape i box model , much like Mop...

Basic edge extrusion method to create the base, then the cut, connect, and chamfer tools to add edges where I need them, and the create poly/vertex tool to add faces where I need them. This works well as I can basically have total control over my edgeloops and poly density.

I start with the head, and work my way down and out. For detail/armor/clothing and other bits, I model them seperately, then either position them where they are supposed to go, or weld them to the main model.

Also...an added thing is to visualize how you plan to get where you're going, including what tools you will need and how you will use them to get there.

This is max specific, as that's all I know at the moment, so bear with me [img]/images/graemlins/smile.gif[/img]

For certain shapes I've found it usefull to start out with a NURBS surface, conform it to the shape needed in a side/front/whatever view by moving the control points/lines around (adding u and v 'lines' of controll points where needed) and then going to the front/side/whatever view to pull the control points into the required 3d shape.
Then a 'convert to poly' and maybe a moment to get rid of some unneeded polys/edges/verts (although if you've done a decent job on the deformations above [and depending on the shape] you might not have to change anything).

This works great for certain shapes, I've found. Faster than creating splines and surfaces, with the added benefit of very easily adding extra loops and very decent control of flow and poly density (depending on how many u/v rows/columns you use).
Creating a pair of skintight-ish pants was done in no time this way....create a sheet, fit it in one view, fit it in another, fine-tune and convert to poly!

Cool techniques, nice to see games people messing around with the non-polygon tools in their apps. Box modeling never dies, does it?

Also, I'm wondering if most of you bother with musculature? I've seen cinematic models where the limbs didn't deviate too much from the basic cylinder. The isoparms were just scaled up and down to create the contour.