Tencent's new AI model, HunyuanWorld-Voyager, creates short 3D-like videos from a single image. Users can control the camera's path to explore the virtual scene. The AI generates both video and depth information, allowing for 3D reconstruction. While not true 3D, the video maintains spatial consistency as the camera moves. Trained on 100,000 video clips, it's limited by its reliance on imitating patterns from its training data, meaning it struggles with unseen situations. Each generated video is short (around two seconds), but multiple clips can be combined for longer sequences.

Prepared by Jonathan Pierce and reviewed by editorial team.