Also look into nvidia's solution... And video files cannot be turned into 3d (without more time than you have to commit). You need a realtime mesh (known vertex coords) to render two versions of a single scene, each render pass has the camera position adjusted to compensate for the human eyes being spread apart.
So its not like we will ever have a plugin to make video pop, but if you do all realtime 3d generative stuff, the option is there. Oh and the projector used needs to be 3d vision capable.
Asus build - intel 2.8GHz T9600, 4GB 800Mhz DDR2, ATI 4670
Rack unit - intel 2.7GHz q6600, 8GB 1066Mhz DDR2, 9800GTX+, intensity HDMI input card