Our AI video capture system captures the timing of animations and rough poses fairly well. 5-10 seconds of animation can be captured in 30 seconds vs 3 hours keyframing the same reference from scratch.
Not as good as a mocap suit, but much more control over the input source and more possibilities with videos from the internet. Noise and foot sliding issues can be 'cleaned' with our graph editor, animation layers and IK/FK system, all in one place.