WorldFoundry
English中文

InSpatio-WorldFM

Method overview: Overview. In the offline stage, a multi-view-consistent model generates plausible observations that provide 3D anchors and reference appearances. In the online stage, frame model performs fast real-time inference.

InSpatio-WorldFM from InSpatio: world and video; Verified.

Edit this pageGitHubMarkdown

InSpatio-WorldFM is a 3D world foundation model for spatial intelligence and frame generation, released with confirmed code and weights (inspatio/WorldFM on Hugging Face, Apache-2.0). In this catalog the integration is verified: the WorldFoundry runner supports the official meta.json demo in the unified environment profile with local WorldFM assets, and emits the official video and panorama artifact structure.

Official paper teaser: Examples of generated worlds across diverse styles, including photorealistic, science-fiction, game-like, and artistic environments. The joystick interface enables real-time interactive exploration with negligible.
Official paper teaser: Examples of generated worlds across diverse styles, including photorealistic, science-fiction, game-like, and artistic environments. The joystick interface enables real-time interactive exploration with negligible.
Method overview: Overview. In the offline stage, a multi-view-consistent model generates plausible observations that provide 3D anchors and reference appearances. In the online stage, frame model performs fast real-time inference.
Method overview: Overview. In the offline stage, a multi-view-consistent model generates plausible observations that provide 3D anchors and reference appearances. In the online stage, frame model performs fast real-time inference.

The recorded paper is arXiv 2603.11911.

Project page: inspatio.github.io/worldfm. Code: inspatio/worldfm. Code: world-fm/worldfm.

Build your run

Pipeline
WorldFMPipelineworldfoundry.pipelines.worldfm.pipeline_worldfm
TaskOfficial Meta Demo official-meta-demo
Environmentworldfoundry-unified-cu128
DeviceCUDA 12.8
Runnerworldfoundry.pipeline
Loadingfrom_pretrained
Invocationnative_or_standard_invocation
worldfoundry-eval run \
  worldfm \
  --pipeline.task-profile official-meta-demo \
  --json

What it is for

worldfm is the recorded interactive world line. Start here when the job is stepping a world with actions.

Outputs cover both video and panorama artifacts, reflecting a world-generation design that renders beyond a single fixed viewport. Architecture internals beyond the official repository and checkpoint pointers are not recorded in this repository.

Catalog tasks include world, video, and world model. Documented outputs include video.

Recorded evaluation entry points include WorldScore (Unified world-generation quality for 3D world output).

Runtime contract

Optional input_path, prompt, interactions, fps, and meta_path; artifacts generated_world and worldfm.mp4.

CLI task profiles: official-meta-demo. WorldFoundry binds WorldFMPipeline.

Family status: Verified.

Recorded official-meta-demo defaults: fps=30.

Drive runs through the official meta.json demo format with local WorldFM assets staged.

Running in WorldFoundry

The catalog launch is worldfoundry-eval run worldfm --pipeline.task-profile official-meta-demo. The command builder above switches --pipeline.task-profile among recorded tasks and still emits worldfoundry-eval run.

It runs in the unified environment worldfoundry-unified-cu128 (Python 3.11, CUDA 12.8).

Recorded weight repositories include inspatio/WorldFM. Prepare the environment and check local assets with bash scripts/inference/prepare_model_infer.sh worldfm. Add --download to fetch missing public checkpoints.

Status: Verified.