His body barely moves at all until it randomly spasms into motion every like 30 seconds. That's not how people, especially Mr talks with his hands, communicates.

I've played a bit with frame interpretation using open source models. With those models, fast movement did seem to trigger some weird artifacts and glitches that were visible at lower speeds or paused.

I guess it's possible that a propietary model is using some other image generation techniques to make those artifacts less noticeable. I think that would still constitute something that should be called out.

If the issue is with interpretation, then we shouldn't see this visual glitch in the original footage.