Was there anything shared on how they achieved these results? One big hurdle for prompt to 3D is usually the super long generation time while they (seem) to do it in seconds.