WorldCrafter Base
WorldCrafter Base from Tencentarc: text-to-video and image-to-video; Verified.
WorldCrafter Base is a world model from Tencentarc. The catalog records it for text-to-video, image-to-video, camera-controlled-video, and world-model.
No official teaser or method figure is checked into this docs tree.
The recorded paper is arXiv 2609.24984.
Project page: drexubery.github.io/WorldCrafter. Code: TencentARC/WorldCrafter.
Build your run
WorldCrafterPipelineworldfoundry.pipelines.worldcrafter.pipeline_worldcrafterworldfoundry.pipelineworldfoundry-eval run \
worldcrafter-base \
--pipeline.task-profile text-to-video \
--jsonWhat it is for
Use worldcrafter-base for camera-conditioned video inference. The recipe binds the recorded camera-control profile.
Catalog tasks include text-to-video, image-to-video, camera-controlled-video, and world-model. Documented modalities: text and image in, video out.
Runtime contract
Optional prompt, image, actions, seed, and frames; artifacts worldcrafter.mp4.
CLI task profiles: text-to-video, image-to-video, camera-controlled-video, and world-model. WorldFoundry binds WorldCrafterPipeline.
Family status: Verified.
Reuses shared Helios transformer, scheduler, latent preparation and VGGT layers.
Model-specific UCPE camera residuals, RepEncoder memory and Fast routing remain explicit.
Running in WorldFoundry
The catalog launch is worldfoundry-eval run worldcrafter-base --pipeline.task-profile text-to-video. 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 TencentARC/WorldCrafter-Base and TencentARC/WorldCrafter-Fast. Prepare the environment and check local assets with bash scripts/inference/prepare_model_infer.sh worldcrafter-base. Add --download to fetch missing public checkpoints.
Status: Verified.

