I was thinking that, too. Even if they use the same model, couldn't they render at lower resolution for TV-- fewer iterations and texture detail and so on?
My understanding is that while you could create lower-detailed textures and map those onto the existing model, if you wanted to reduce the poly count you'd need to create a whole new model. You can't just say "hey, magically delete some random polys from this".
Not exactly. There are ways.
One trick is to essentially strip away layers of greeblies (which typically amount to the lion's share of any hight poly model of this sort) all down one side and just never "shoot" it from that angle. Same for loading textures. No point burning render time on something that's not going to be visible anyway.
Another is to render those greeblies as a separate file and just apply them as a texture/bump-map.
So long as the camera doesn't get too close, it shouldn't be all that noticeable.
As I said though, it's not just about polys, but texture-res and multi-sample passes as well. Plus of course the final pre-composite render for a TV show isn't likely to be much more than 1080p or 2k max. For a modern movie you're probably looking as something more along the lines of 8-16k which is obviously several orders of magnitude greater.
