Post
2565
Just tried tencent/HY-World-2.0 โ a multimodal world model that takes in text or a single image and generates editable 3D scenes.
Unlike Google's Genie and HY-World 1.5, v2.0 generates engine-ready 3D content:
๐ฎ Direct import into Unreal Engine and Unity โ no format wrangling
๐ง Supports multiple 3D asset formats: Mesh, 3DGS, point cloud, etc.
โ๏ธ Fully editable โ not a baked video, but actual geometry you can modify
๐ค Also usable for embodied simulation environments
Basically: from "AI generates a world you can look at" โ "AI generates a world you can ship."
Unlike Google's Genie and HY-World 1.5, v2.0 generates engine-ready 3D content:
๐ฎ Direct import into Unreal Engine and Unity โ no format wrangling
๐ง Supports multiple 3D asset formats: Mesh, 3DGS, point cloud, etc.
โ๏ธ Fully editable โ not a baked video, but actual geometry you can modify
๐ค Also usable for embodied simulation environments
Basically: from "AI generates a world you can look at" โ "AI generates a world you can ship."
