We are trying to create a fighting game. And we are going for every kind of moves and weird moves too.

And would like everybody's opinion that, should we have the models in a single mesh or with multiple meshes. Say upperarm, lowerarm, hand, thigh, calf, boots, torso and head. Both methods have their own advantages and dis-advantages.

We would like your opinion as to what is the Ideal method one should adopt.

I would ever prever a single mesh, cause it has nice phong sahding, and no that hard borders.

But our problem with single mesh is that, the animations we are doing, dont look right, and polygons come out as weird. There is very much streching, bad looking polygons and as the model is rotated in possible angles, which doesnt gives a decent look for the model.

Only a dissected mesh solves problem relating to strechicg and not being subjected to weird polygons. But it has its own issues.

Can there be any good choice, as we are going for bones animations and there is the choice of one bone - one vertex.

its a little hard to understand by what you mean by "bad looking polygons". I assume you are talking about the stretching of polygons around the joint areas, as there is only single bone per vertex (elbows, knees, hip)

There is a very careful way to setup these joint areas to minimize or even prevent the polygons stretching. I dont have the picture, but im sure someone on the forum should have it. In short, it shows that adding a few extra polygons around the joints that are specifically made to stretch can minimize the problem of a single face(s) stretching too far during an obtuse rotation.

I wouldnt create the model out of seperate meshes reguardless. Thegamemaker's comment is absolutly right. An unclosed mesh dosent have uniform lighting since it is actually seperate peices, and can look downright ugly.

You may have to rework your joints to accomadate stretching, and start making any later models with the mindset of building the joint polys to stretch.

EDIT:

found it:

The example you have here has the option of assigning more the one bone per vertex. Our problem is that every vertex can have one bone. It is not only the stretching problem but, we are rotating bones in great angles and thus polygons pop up at different angles.

We are going for every possible move a player can perform. And the body can twist and turn at very many angles. We are yet to find a model that would suit our needs.

If any body has an example model that they think, can be used to perform any kind of moves then please let us have an example, so that we can really decide that if is it really possible to have a model that can perform any kind of move without much degradation.

We are doing sweep kicks, hand-stand attacks, jump kicks, sommersaults, flips, and many other unique moves with the models. And have not found any ultimate single mesh model that would fit our animation requirements.

The middle animation of the picture above dosent use multiple bone per vertex, I should have made it clearer.

Well it sounds like your animations are going to be very expressive in range of motion, so the only thing I could say is that the actual time of the stretching should be made minimal. Since its a fighting game, the moves shouldn't take inordinate amounts of time on screen. If a character is to bend his torso 180 degrees backward, then this should be a quick animation, not long enough to expose the limitation of MED at the moment. I assume the character's standing animations dont call for extreme rotation, as in my opinion that would be out of place even in your type of fighting game.

If your going to use bones, stretching will be inescapable even with careful modelling of the trouble areas. However, you can combine animations, rotate heads and eyeballs toward the enemies, attach armor and weapons with minimal effort compared to a workaround with vertex animation and scale limbs realtime. Bone animation should be chosen if youd like to do these things.

Vertex animation however lets you hand animate any and all problem areas, and generally vertex animation looks better if the animator animates in layers of detail (like any professional hand animator). You can make muscles buldge, skin wrinkle and stretch with vertex animation. If you wanted to do that with bones, you would have to attach bones to every desired vertex. If a torso rotation looks too wooden, the animator can stretch and pull edges to curve out rotations.

Personally I wouldnt choose vertex animation because its impossible (or not worth the effort) to rotate limbs toward targets (head rotation independent of torso). You can't combine two animations (torso is punching, legs are running) and change these in real time. If you wanted the head to rotate with vertex animation, you would have to make the head a totally seperate model and attach it to the body (vec_set, vec_for_vertex), and this would show some obvious lighting inconsistancies with the head and body.

Degradation or unadaptability, take your pick.

As you have stated, for the reasons and advantages of bones manipulation through script, we have opted for Bones animation. The main problem areas are shoulder joint and thigh joint. which have to have extreme manipulation and are very problemetic at present.