Open Frooxius opened 1 year ago
Related thread with info, comparisons and feedback: https://github.com/Yellow-Dog-Man/Resonite-Issues/issues/130
Will this also incl #15 which is hand grip finger curl limit? So you can set how far the fingerss can curl inward when gripping button is pressed this fixes so fingers don't come out the back of the hand on some avatars
Would a new rig need to be made for avatars or will it be backwards compatible with the current rig naming etc?
Related thread with info, comparisons and feedback: #130
Thank you!
I'm going to do more comparisons in here from now on when I have the time to do so!
With this issue, would it be also ok to generalize and expose IK components as well? This would cut back on requirements for the issue somewhat, and also allow folks to make their own custom iks a bit easier (Taurs, Multi-peds, other systems outside of arms)
Twist bone support is also something I realize shouldn't be missed as well.
Will this also incl #15 which is hand grip finger curl limit? So you can set how far the fingerss can curl inward when gripping button is pressed this fixes so fingers don't come out the back of the hand on some avatars
No, that's not part of this. Fingers are a separate system.
With this issue, would it be also ok to generalize and expose IK components as well? This would cut back on requirements for the issue somewhat, and also allow folks to make their own custom iks a bit easier (Taurs, Multi-peds, other systems outside of arms)
Twist bone support is also something I realize shouldn't be missed as well.
Adding onto this, I do wonder if manually assigning rigs could be a thing too because for example: unity has this:
So while you should be using proper bone naming, if you don't you won't be completely screwed and you could reinitialize the IK by setting up some of the bones manually. Sometimes avatars for other platforms have some issues in Resonite due to the person naming bones improperly, and instead of having to reimport the whole thing it would definitely be nice to be able to do that.
With this issue, would it be also ok to generalize and expose IK components as well? This would cut back on requirements for the issue somewhat, and also allow folks to make their own custom iks a bit easier (Taurs, Multi-peds, other systems outside of arms)
Twist bone support is also something I realize shouldn't be missed as well.
This has been touched on in the initial post. The goal is to generalize things where possible, but it will be avoided if it comes at the cost of performance, the IK behaving well for humanoids (which is the primary goal).
Ah okay cool! Sorry missed that bit about the generalized components!
I've created a discussion for this, to keep things better organized. Please direct feedback, thoughts, wishlights and other stuff here: https://github.com/Yellow-Dog-Man/Resonite-Issues/discussions/616
Wondering cause I wanted to make a new issue for it but I don't know if a better solution is in the plans for the new IK, but would avatar setup/post setup still behave about the same? I wanted to create a suggestion/feature request to add some kind of button that lets you preview an avatar's viewpoint during the setup process to make placing the eyes a bit easier
The avatar creator itself is due for an overhaul/rewrite, but that is a separate issue/work item from the IK itself, @Dusty-Sprinkles.
I talk about it a bit here: https://github.com/Yellow-Dog-Man/Resonite-Issues/issues/20#issuecomment-2078294380
Is your feature request related to a problem? Please describe.
The current system for full body IK for avatars is a heavily modified port of FinalIK. As such, this system has a number of issues:
Describe the solution you'd like
After digging around the FinalIK-based code for a long time, I believe the best and overall easiest solution is to take full ownership and build our own in-house full body IK system from scratch, using the accumulated knowledge of IK systems, issues that users have with various avatars and full body tracking setups.
This way we solve all the outlined problems at once and get a good, clean codebase designed to integrate well into FrooxEngine to work with into the future.
There are number of design goals for this IK:
There are some smaller ones, but these are the key points. I'll add more info as time goes when necessary too.
Describe alternatives you've considered
We considered just keeping FinalIK and making incremental improvements to it, but this would only compound the problems and it's increasingly harder to work with. To solve some of the problems we were facing, we'd have to redesign how FinalIK is architect-ed in the first place, essentially rewriting it with a custom system anyways.
Alternatively we could also look for other 3rd party systems, but generally we'd run into the same issue there - they do not integrate super well with FrooxEngine and its needs. The IK is one of the key systems for platform like Resonite and something that's used daily for people to express themselves and inhabit their avatars, so having an in-house system tailored for the needs of the platform and with codebase we fully own and control is going to be highly beneficial in the long run.
Additional Context
As this system will be developed, there will be a period of overlap, where both FinalIK and our custom ones exist. This will allow for comparisons and easy testing as the custom system is iterated on.
Once our IK achieves feature parity and behaves well enough for most avatars, we'll perform switch and make it the new default. Following which we'll start phasing out FinalIK - converting existing avatars to the new system and eventually removing FinalIK completely.
Additionally I expect as components and algorithms are written for this IK, we'll trickle some of them in as individual components too - e.g. trigonometric and FABRIK solvers, that can be used for any chains and/or to build custom systems, so there will be some nice goodies to play with along the way too.