Google releases Genie, a base-world model trained on Internet video with a parameter scale of 11B, which can generate a myriad of motion-controllable interactive environments from composite images, photographs, and even sketches.Genie can be cued with images it has never seen before, the paper's address:https://arxiv.org/abs/2402.15391
Google Releases 11B Base World Model Generating Interactive Virtual Worlds
Previous: 南京大学开设AI通识核心课程体系
Next: Pika推出“对口型”功能