Cage,

You were asking about the bias files we were playing with. Here is an animation of two standard Antonia figures. Both have 50% Sexy Walk applied. The one on your left is without the bias, the one on the right is with bias applied. As you can see there is still a jerk in the left knee, but with 50% Sexy Walk there is an even worse jerk in the right knee for both figures (Les puts head in hands and sobs).

Like I said, I'm struggling to remember details. I will give you more info, when and if my memory kicks into gear. The basic process of producing the above animation was to inject some slaving code, then inject bias settings. Then run the WD, then turn the ERC on.

The ERC version would be useless as input to the WD, because the WD does not understand ERC. I think the idea was, that once I had the bias settings figured out, your script could then apply the settings to a pose file, and that pose could be used as a new walk blend. There again, I may have got that bit completely wrong. I'm still struggling to remember.

The idea shows some promise, the biased walk does look a bit better IMO, but is still a long way off being good. The real holy grail is to produce a walk blend that outputs a good walk, without the need for further tweaking.

P.S. I suspect the knee twitch may be due the the hip being too low in that frame, and IK is throwing the knee to the side because it does not understand which way to bend it. That's one of the big problems. We can't just come up with walk poses that look good when applied as animated poses, we also have to take account of how IK will handle the pose when it is used as a walk blend in the WD. I think that hip altitude, and possibly residual thigh and shin rotations may be the key factors here.