Back to news
researchXiaomi2026-11-04

Xiaomi reveals WorldCast on November 4, adding 3D camera motion generation from still photos on REDMI K100

WorldCast interprets depth structure to reconstruct a 3D scene from a flat photo and virtualises a realistic camera trajectory, keeping subject appearance consistent during motion.

Reported on November 4, the Xiaomi REDMI K100 phone was formally released on October 27, and according to Xiaomi's official technology account the device adds an AI-based 3D camera motion feature to imaging. Selecting a still photo from the gallery lets AI generate a large-angle 3D camera motion video, with the camera able to orbit, push in, pull out and tilt while the subject's facial features, hairstyle and clothing stay consistent.

Supporting this feature is WorldCast, a visual generation framework brought by a recent Xiaomi research paper. It interprets the depth structure of an image to reconstruct a flat photo as a three-dimensional scene and virtualises a realistic camera trajectory within it, so the camera moves as if actually filmed and subject appearance stays consistent throughout the motion. Xiaomi said WorldCast is not merely an extra trick for the REDMI K100 gallery but an accumulation of method, dataset and evaluation methodology, with the team going on to explore more complex camera motion forms, longer continuous generation and controllable 4D video generation.

Binding a consumer phone imaging feature to a reproducible visual generation framework is a common route for device makers to turn research results into product features, and it also brings controllable 4D video generation closer to mass-market use cases.

WorldCast小米视觉生成4D 视频