New AI model turns photos into explorable 3D worlds, with caveats
SMRTR summary
Tencent's new Voyager AI model transforms photos into navigable 3D environments by building on its Hunyuan ecosystem and using an automated data pipeline trained on 100,000 video clips. While achieving high scores on the WorldScore benchmark, the model requires substantial computing power (60-80GB GPU memory), has licensing restrictions in certain regions, and faces challenges for real-time applications despite supporting multi-GPU processing for faster results.
SMRTR provides this summary for quick context. The original article belongs to Ars Technica.
Read the original article