The engine IMVU uses is equivalent to idTech II (Quake 2, maybe early idTech III) in capability -
Pixomatic Rendering Technology. In theory the numbers - mesh density, texture sizes and so on - are supposed to be relatively low, it's more or less akin to Quake III limits for scenes and items that drop into it, however characters models are c.2500 tris in default dress but can go into the as high as 100k (usually because 99.99% of users making content for IMVU don't understand/don't care about proper game development principles and optimisation techniques).
The world is modelled and lit semi-dynamically - a limited number of light entities (no more than three are recommended) are place in to a scene which vertex shade meshes relative to their position. No lightmaps or other 'artistic' shadowing so everything either has to be vertex painted at the mesh level or baked into the texture (as we would do with any other type of model).
Characters don't 'walk' like other first/third person games, they move using location markers, you click one and the avatar moves to it. Character actions and animations are then carried out relative to the spot the avatar is on so they can 'move' but reset back to the starting spot (in instances of animated short walk sequences), or move around the spot (looped sequences).
This is an
old presentation on IMVU but gives a better idea of what the 'game' is about (the name is pronounced "im vu" rather than "i m v u" and whilst it doesn't specifically mean anything in terms of it being an acronym for something - a
common misconception is that it means "
Instant Messaging Virtual Universe", it doesn't - the term does correlate to shorthand references for "IM" being "Instant Messaging", which is what the program actually is, with "VU" ("view") similarly correlating to "3D View" - the medium through which the program functions).