Hi everyone!
I've read this interesting discussion
Retro 3D Art FAQ - Everything you need to know to create PS1/N64/Dreamcast/etc. 3D art! — polycount and I'm interested in the
"Character rigging via Parenting" part
: " In the early days of realtime 3D character bodies were often split up so that each limb was a separate mesh with its own pivot". I'm studying how old games are made. This technique is common among retro console games, but I was wondering if you know some old PC games using this technique too, like Ečstatica, without any skinned mesh at all?
Thank you for your answers!
Replies
https://www.youtube.com/watch?v=UDm6F463QMI
A lot the early 3D games from the original PlayStation used 1:1 skinning.
Final Fantasy 7 (36:00)
https://youtu.be/rwiVRAOJPDs?si=CWrYEs4vhrwKfOy2&t=2158
Resident Evil (3:20)
https://www.youtube.com/watch?v=E0U7IslEpIg&t=105s
Dino Crisis used 1:1 skinning, but they didn't separate the meshes into piece. Still each vertex was weighted 100% to a bone
https://www.youtube.com/watch?v=-kNTd_r5zdI
Quake 3 took a hybrid approach with the torso, head and legs being separate skinned meshes.
quake and quake2 were just canned vertex animation iirc
I think it'll mostly be games that also launched on ps1 / equivalents that did this tbh.
virtua fighter and tekken did it iirc
I think segmented mesh animation was a performance limitation, not stylistic choice, so since the target PCs could handle skeletal animation already, the programmers would go with that. That's why we can't find many PC games with this type of character animation.
That's not exactly true. In QIIIA the elements with smooth deformations (torso and legs) are not skinned meshes (as in, driven by a skeleton) ; they are purely vertex animations. They can of course be authored as skinned skeletal assets, but they are not exported as such. So in the end the engine simply handles the relative rotation between head, torso, and legs based on player input. No bones involved, and I believe the head part didn't support animation at all, only reacting to mouselook but not deforming.
To the OP : overall there's much more to it than just segmented VS smooth. Also and FYI, "rigging" actually means the work done at the authoring stage (in Max/Maya/Blender/C4D), consisting of building complex systems to drive bones with special controllers (for instance : setting up a lookat constraint to direct the gaze of a character by simply moving a point in space, or limbs IK to pose arms and legs). So going back to the QIIIA example, a QIIIA character could be *authored* in Max/Maya using a very complex rig with many bones and advances constraints (Magdalena ! ) ; yet the game would see none of that since the export for a bodypart simply consists of a raw matrix of vertex positions changing over time.
Inversely, just because a game has segmented characters doesn't tell you much about what the engine actually does. A visually segmented model could be simple child/parent parenting, or it could be geo weigthed to a skeleton but only using 0s and 1s.
I would say that if an old game has very crude and typical "robot-like" animations (think : early 3D arcade games like Virtua Cop1), then the assets are probably parented without the use of a skeleton, and were probably animated by simply rotating sub-objects without even using a skeleton. Of course there's also the possibility that such robot-like animations are simply the result of the game being done by a very small team without human or tech support for actual rigging. And then there's the case of Mocap, allowing for something like Virtua Fighter - not necessarily requiring *any* advanced rigging for the creation of animations, since they come already done. Which explains why the characters in early VF games move very convincingly ... but their claw hands move in a very crude manner - because there's no mocap for the hands.
If anything this explains why 3DSMax was so popular in gamedev for a while, as despite being inferior to Maya for fully custom rigs it came with Character Studio, which allowed for an unusual but very intuitive way of setting up an humanoid rig and had clever animation controls that allowed to animate without having to interact with typical animation curves ; and also, it provided a unified exchange format for mocap data. So basically the best of both worlds, turbocharging the creation of third person games and convincing game characters in general.
I believe consoles tend to have a more varied history of technical implementations of 3d characters because of fighting games and third person games being more of a consoles and arcades thing. A look at game lineups for 1995/1996 shows that PC games had barely any 3D characters at the time - 3DFX was only founded in 1994, and the Voodoo2 only came out in 1998.
He mentions only briefly about their particular animation system for q3arena at the very beginning of the book.
RIP Paul.
it's amazing how little has really changed in the last 20 years if you ignore the specifics of the tools.
https://www.youtube.com/watch?v=urIqLheLwtU