Tencent has released version 1.1 of its Hunyuan World Model, codenamed WorldMirror, adding support for multi-view and video inputs while enabling single-GPU deployment for real-time 3D scene creation. The company said the update allows users to generate 3D worlds in seconds.
As a unified feedforward 3D reconstruction model, WorldMirror 1.1 expands beyond version 1.0’s text and single-image inputs, now supporting multimodal priors such as camera and depth data with end-to-end 3D output. The model is fully open source on GitHub, with a Hugging Face Space available for online testing. [TechNode reporting]
Related
