Sliding scale of avi detail
I have some thoughts on this. Following hf’s general design direction, I think a clearly defined, sliding scale of avatar quality is required - from being reduced to a single coloured voxel right up to a detailed close-up of an eye or finger when, for example, adjusting attachments. Following this idea through to the n’th, I see a requirement for a hugely highly detailed avatar - e.g. multiple facial muscles, two roll-bones per limb, the complete set of 33 vertebrae spine joints, fingers with 3 rotation axes, etc, etc - turn it all up to 11! The level of detail described by the avatar would obviously be dependant on factors like available bandwidth, graphical / processing power of client platform, availability of contributed devices etc etc. I think this idea fits hf’s high fidelity goals and closely matches the scenario of low-end, greyscale and fuzzy avis from the public terminals in the book that doesn’t need mentioning
Animation retargetting woes - 3 avi solution?
The problem, of course, with having any sort of variability in the joint structure is the subsequent fitting of animation sequences. A certain amount of automatic animation retargetting is possible, so perhaps work could be done producing two fairly standard (male and female) avatars whose proportions are loosely constrained to actual human proportional limits. The task of automatically retargetting ‘standard format’ animations would be easier than for a more general case, as there would not be such extreme boundary conditions to deal with. However, having had a good crack at implementing automatic retargetting in the past, I can confirm it’s a far from trivial task! (see here)
A third, far more customisable avatar for people who want to take avi customisation further could be also offered. This avi would be highly configurable, so people choosing to purchase a custom avi - an animal avi, for example - would probably pay more because by necessity it would include it’s own, unique bone structure and set of animations.
So, in defining three (or even more) classes of avatar, we could go some way towards catering for the hugely broad range of technical ability of users in avatar customisation.