And we don’t know if there is any “inpainting” like the real-time chunking from pi? I would be surprised if just a VLA can produce something this smooth - must be much higher frequency than what VLA currently run at, no?