A "coffee or tea" length tutorial on how to utilize the very powerful (and FREE) curve editor in Blender to cleanup and refine mocap before you bring it into iClone.  Blender can be intimidating so I think the best way to utilize it is to learn simple specific tasks - like editing mocap data!  I always recommend watching a whole tutorial once and then follow along on the second view.  Good Luck!  

[YouTube]https://www.youtube.com/watch?v=5hRz60MaQIk[/YouTube]

Hey Mark ....Great stuff!
To see a walk-through for specific 'tasks' like this, complete with the 'new user' in mind is terrific.
The *Smooth Keys* function - very useful - "multiple iterations" of it - great!
Yes - I experience those *rotation issues* (spasms)...

It would be great if Blender could somehow automatically determine what's 'rotation' vs what's 'location' data.
Imagine, *if* it could 'sense' the smallest motions (the high frequency jitters) - then with a slider - you could progressively reduce the amplitude of the 'jitter-data curves'!

Others have mentioned (and I also find) that the legs seem to suffer with mocap tracking errors the most.
I've come to the conclusion recently (at least for now), that, for movie making:
● Wide character shots (where legs are seen) and where there's walking/running - are best handled with stock motion files, while:
● Mid shots (torso and arm action, dialogue) are most suitable for mocap, and
● Closeups (facial expression and dialogue) are best executed by manual face animation or with *live actors* (which would yield the very best performance). (based on not having access to facial animation software). This is just my take on things ....

Of course *mid-shots* or *wide-shots* that have dialogue could still receive manual face animation (or even use live actors), but *generally speaking* the above ●'s are optimal for their intended use - (the way I see it - for my use atm).
- Meaning that leg mocap issues (which are really common) - wouldn't have to be 'dealt with' at all.

But I'm digressing a bit here....
So really - this *Blender* approach to clean-up, I can see being very useful for arms, hands, neck and head (for my particular approach).

But this is a very helpful video - no matter which body sections need cleaning up, and no matter which mocap hardware/software a person uses ....  so thnx heaps for this Mark 

--------------------------------------------------------------------------------------------------------------------------------------------------------

Please be patient with me ..... I don't always 'get it' the first time 'round - not even the 2nd time!   - yikes! ... 

● MSI GT72VR Laptop, i7 7700HQ 4-Core 3.8 GHz 16GB RAM; Nvidia 1070, 8GB Vram ● iClone-7.93 ● 3DXChange Pipeline 7.81 ● CC-3 Pipeline 3.44 ● Live Face ● HeadShot ● Brekel Pro-Body ● Popcorn FX ● iRAY ● Kinect V2 ● DaVinci Resolve17 ● Mixcraft 8.1

I would say your conclusions are only relevant to the type/system of mocap you are using.  High end mocap systems (like the PN) track *everything* well, and are suitable for just about any motion you can think of.  Indeed, how do you think those canned walking/running motions you are using were produced?  I can promise you they at least started off as mocap (with perhaps some hand refinement).  You can really tell good mocap, as it is almost impossible to hand animate things as well (almost all of the stuff RL provides for, say, Heidi and Mason almost surely started off that way).

The cheaper the mocap, the more it will have problems with things like leg and body positions.  You could even expand your list to say things like:

Cheap mocap works best if the body stays even with the capture device (assuming you are using a visual capture like the Kinect) and when the limbs do not occlude with the body (ditto).

Cheap mocap works best with *slower* movements -- the faster the movement, the more trouble it will be (so if you need to do fast movements, slow them down during capture and speed them up afterwards).

The Leap controller has some of those same limitations, for the same reasons (you can't occlude one hand with the other, you need to be careful about very fast motions). 

Higher end systems don't worry about such things, but it all comes down to you get what you pay for.

---

Alienware Aurora R16, Win 11, i9-149000KF, 3.20GHz CPU, 64GB RAM, RTX 4090 (24GB), Samsung 870 Pro 8TB, Gen3 MVNe M-2 SSD, 4TBx2, 39" Alienware Widescreen Monitor
Mike "ex-genius" Kelley

Yes Mike, I'm just trying to identify the strengths of what I've got - limited though they are ....
Working with limitations is a pain, but either I fight it (which I can't win) - or I go with it, (which is winnable) - so yeah ... 

>>> "... Cheap mocap works best with *slower* movements -- the faster the movement, the more trouble it will be (so if you need to do fast movements, slow them down during capture and speed them up afterwards)... " <<<  - Good to know Mike .... thnx ...

--------------------------------------------------------------------------------------------------------------------------------------------------------

Please be patient with me ..... I don't always 'get it' the first time 'round - not even the 2nd time!   - yikes! ... 

● MSI GT72VR Laptop, i7 7700HQ 4-Core 3.8 GHz 16GB RAM; Nvidia 1070, 8GB Vram ● iClone-7.93 ● 3DXChange Pipeline 7.81 ● CC-3 Pipeline 3.44 ● Live Face ● HeadShot ● Brekel Pro-Body ● Popcorn FX ● iRAY ● Kinect V2 ● DaVinci Resolve17 ● Mixcraft 8.1

Hey Steve,
Thanks for the appreciation!  For years I've only used Blender for features I didn't have available elsewhere - like smoke and water simulations... trouble is, I only learned how to do those specific tasks and have yet to get my head around it.  I have to say, the more I learn the more I like it... it's different, so it feels awkward but that's because we are used to 3D software designed by engineers - Blender's workflow actually feels more like it's been conceived by artists/animators and from what I've experienced on the upcoming v2.8 it's about to make a 'leap' in ergonomics, function and real time rendering.  For instance, once comfortable with the F-curves you will find that 'human' solving is better than any 'algorithm' solving with regard to "what" needs cleaning.  So, selectively deleting or smoothing keyframes whilst leaving the key motor movements alone is always going to yield better results than a "one button" process.  WebAnimate is a great example of 're-solving' the mocap to help cleanup the motion - but 'pre-cleaning' the mocap manually in Blender or other, will only give WebAnimate a better starting point for the best possible result.  I've recently experienced mocap from a 10 camera Vicon system (Hollywood grade) and I can testify that there are many hours of 'manual' tweaks needed for every capture (not to mention ALL the hand/finger animation!).  Bottom line... if Hollywood can't afford perfect "out of the box" mocap then we shouldn't expect it either.  Food for thought... pro 3D software like Blender can 'copy and paste' animation curves - who's to stop you from using "prefab" walking animations "Frankensteined" to your own upper body animations (providing the skeleton rigs are identical)?  Or similarly... combine the best of 2 or 3 kinect style mocaps?  We can all "copy and paste" ;-)