When animating w/endomorphs, obviously, the more points there are in the model, the more work the computer has to do to morph the model.  This is true whether animating in LW or working w/another program like TAFA.  So here's a technique for lowering the point count while animating.  It can definitely increase the playback speed of your animation and probably has many applications.
 
Let's say we have a full body character and we are going to do some facial animation so we don't need the body pgons while we're animating.  Cut the body pgons from the model and put them in a separate layer.  Save this model as modelname_split.lwo.  Now delete the body from the model and save it as model_HeadOnly.lwo (or you might be able to just load the layer containing the head, depending on which app you are working with). Now do your morph-based animating.  If you are working w/Morphmixer, you can actually just apply the morph mixer files to the original model, since they are just text files which reference the morph names.  If you need the full model for final output of the animation files (ie. for motion designer output), after animating and testing with the head-only model, you can replace the head with the whole body model and then export the .mdd data.  But that's not all!   If you need to add or modify morphs in the model, you can make the changes to the head only model, then open modelname_split.lwo, delete the layer with the head, then copy and paste the new head into the layer with the rest of the body. merge points, and the new morph targets will come with it.  This is an advanced technique, so be aware that you can easily make modifications to the head-only model, (for example, move the points that will later merge back into the body mesh), that can cause big problems.
 
Note, this example was tested on a model in which all of the morphs are on the head pgons of the model.  If you have morphs for things like the fingers, you will need to test this further than I have (though I think it will still work).

TAFA currently does a certain amount of processing even on surfaces that are not visible.  This is a side effect of which kinds of indices I have into the various levels of model information.  With the current structures, it would actually be slower to try to avoid the calculations.  However, I know the kinds of extra information I need, and these structures will be changing for the 1.5 build anyway, so v1.5 should almost completely remove the need for this kind of proxy.
 
But, it is very useful with version 1.0 or 1.1 of TAFA, especially for detailed models.  Thanks for the writeup!

Kinda depends on what you mean by "better"  
 
One of the things I'd like to do is add different ways to select sets of polygons -- pick sets, etc -- instead of just surfaces.  The trick is how to generically refer to these different things when I'm having to deal with three or four object file formats.  This would be the most affected by data structures, but will probably have to wait for version 2.0.
 
The other possibility is to actually give you pan/zoom controls instead of just rotation controls.  Strictly speaking, this isn't that hard, at least for the main head.  But I'd have to pick out a control method that wasn't already in use, and work out whether or not a pan or zoom is "global" -- applied equally to all models -- or "local".  I tend to think you'd want it global.  If, for whatever reason, you needed to manually adjust which part of the model you're looking at, you probably want to adjust it everywhere, and don't want to have to manually adjust all hundred of your morphs and all of your palettes...  This also will probably wait until v2.0...

Quote:

 
I'd vote for the pan/zoom controls, especially if it isn't too hard to do and wouldn't slow anything down.  At this point I wouldn't care if it was global or not, as I find myself mostly working with the main window.  But I definitely see your point about the possibility of wanting to zoom in on all the palettes.  Perhaps you could add a switch to make it global or local.
 
M