When something on that scale is not data backed, it’s called vibes and feelings.

Makes sense. My RTX 5090 can locally take a text prompt, generate a 2D image from that, feed it into a 3D generation model, rig up the animations, and make me assets that are better than a lot of video games on the market. Not Cyberpunk 2077 level m, sure, but pretty good and this is just what I can clobber together in my free time from open models.

Are the results in video generation precise enough to remove the need for 3D modeling?