WorldFoundry
English中文

ReCamMaster

Method overview: Overview of ReCamMaster. Left: The training pipeline of ReCamMaster. A latent diffusion model is optimized to reconstruct the target video V t V_{t} , conditioned on the source video V s V_{s} , target camera pose c.

ReCamMaster from Kuaishou (Kling Team): camera-controlled video rerendering; Verified.

Edit this pageGitHubMarkdown

ReCamMaster re-renders an existing video along a new camera trajectory — camera-controlled generative rendering from a single video. In this catalog the in-tree pipeline, operator, and synthesis runtime live under the Kling namespace, and the runner is verified for the official camera-trajectory video-to-video demo with local ReCamMaster/Wan assets.

Method overview: Overview of ReCamMaster. Left: The training pipeline of ReCamMaster. A latent diffusion model is optimized to reconstruct the target video V t V_{t} , conditioned on the source video V s V_{s} , target camera pose c.
Method overview: Overview of ReCamMaster. Left: The training pipeline of ReCamMaster. A latent diffusion model is optimized to reconstruct the target video V t V_{t} , conditioned on the source video V s V_{s} , target camera pose c.

Project page: jianhongbai.github.io/ReCamMaster. Code: KlingAIResearch/ReCamMaster.

Build your run

Pipeline
ReCamMasterPipelineworldfoundry.pipelines.kling.pipeline_recammaster
TaskDefault default
Environmentworldfoundry-unified-cu128
DeviceCUDA 12.8
Runnerworldfoundry.pipeline
Loadingfrom_pretrained
Invocationnative_or_standard_invocation
worldfoundry-eval run \
  recammaster \
  --pipeline.task-profile default \
  --json

What it is for

Use recammaster for camera-conditioned video inference. The recipe binds the recorded camera-control profile.

Per the paper (arXiv 2503.11647), ReCamMaster conditions a pre-trained text-to-video model (Wan-family in the open release) on the source video through frame-dimension conditioning, letting camera trajectories be swapped at generation time. Runtime execution requires local Wan2.1, the ReCamMaster checkpoint, torchvision, and CUDA assets. It is published by Kuaishou (Kling Team).

Catalog tasks include camera-controlled video rerendering. Catalog aliases include Camera-Controlled Generative Rendering from A Single Video.

Recorded evaluation entry points include CameraBench (Directly evaluates camera-trajectory control quality).

Runtime contract

Optional prompt, input_path, frames, fps, and steps; artifacts generated_video, generated_4d_scene, and recammaster.mp4.

CLI task profiles: default. WorldFoundry binds ReCamMasterPipeline.

Family status: Verified.

Recorded default defaults: frames=81, fps=8.

The open-source reference runtime is checkpoint/GPU gated — no empty runner shim is registered.

Running in WorldFoundry

Launch worldfoundry-eval run recammaster --pipeline.task-profile default to follow the camera path. 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 KlingTeam/ReCamMaster-Wan2.1 and Wan-AI/Wan2.1-T2V-1.3B. Prepare the environment and check local assets with bash scripts/inference/prepare_model_infer.sh recammaster. Add --download to fetch missing public checkpoints.

Status: Verified.