REPROCESSING AGENT
Turn evidence into modeling context
Validate assets, align motion conventions when trajectories are present, and parse visual evidence into entities, relations and cameras.

Program geometry, motion and viewpoints. Generate high-fidelity robot videos through a collaborative agent workflow.
01 / PIPELINE · BWM-FORGE
From diverse evidence to an executable scene and a high-fidelity video. Select an agent in the collaboration map to explore its internal workflow.
A reference image anchors the scene; robot assets and the task guide motion generation.
First frame + taskAdd Parquet or NPZ motion data to specify joint or end-effector motion and timing.
First frame + robot trajectoryUse demonstrated actions and object-state changes to reconstruct the task for the target robot.
Human demonstration videoMULTI-AGENT COLLABORATION
Four agent roles coordinate input processing, construction, review and rendering. Select an agent to explore its internal workflow.
REPROCESSING AGENT
Validate assets, align motion conventions when trajectories are present, and parse visual evidence into entities, relations and cameras.
BUILDING AGENT
Scene planning and refinement lead to robot import, composition and motion validation. Scene and composition reviews feed corrections back to the relevant steps.
RENDERING AGENT
The Proxy Video guides geometry, motion and timing. The reference image and prompt guide appearance. Scores and user feedback support iterative refinement.
REVIEWING AGENT
Collect visual evidence, revisit numerical and physical evidence in execution logs, and correct localized defects. Preserve the review record and unresolved issues for user feedback.
02 / CASES
Fourteen tasks across three input modes. Choose a result to explore its available references, modeling previews, Proxy Video and render versions.
TYPE 0 · FIRST FRAME
FROM EVIDENCE TO RESULT
TYPE 0 · FIRST FRAME
FROM EVIDENCE TO RESULT
TYPE 0 · FIRST FRAME
FROM EVIDENCE TO RESULT
TYPE 1 · TRAJECTORY
FROM EVIDENCE TO RESULT
TYPE 1 · TRAJECTORY
FROM EVIDENCE TO RESULT
TYPE 1 · TRAJECTORY
FROM EVIDENCE TO RESULT
TYPE 1 · TRAJECTORY
FROM EVIDENCE TO RESULT
TYPE 1 · TRAJECTORY
FROM EVIDENCE TO RESULT
TYPE 1 · TRAJECTORY
FROM EVIDENCE TO RESULT
TYPE 2 · DEMONSTRATION
FROM EVIDENCE TO RESULT
TYPE 2 · DEMONSTRATION
FROM EVIDENCE TO RESULT
TYPE 2 · DEMONSTRATION
FROM EVIDENCE TO RESULT
TYPE 2 · DEMONSTRATION
FROM EVIDENCE TO RESULT
TYPE 2 · DEMONSTRATION
FROM EVIDENCE TO RESULT
03 / EVALUATION
Seven objective metrics measure visual quality and motion. Three vision-language assessments check realism, coherence and task consistency.
Normalized to 0–100 · higher is better
Clarity of image details and edges.
Penalizes excessively dark or bright pixels.
Color consistency between adjacent frames.
Penalizes abrupt brightness and appearance changes.
Continuity of optical flow and motion.
Scene consistency without unwanted camera drift.
Compare motion direction, intensity and timing against the Proxy Video.
Weights follow the current implementation.
Reference + temporal evidence + task
Materials, lighting, shadows, geometry and generation artifacts.
Robot structure, contact, spatial relationships and visible continuity.
Target identity, required actions, event order and final state.
Judgments use the reference image, chronological video frames and task instructions.
S = 0.80 Sobj + 0.20 SVLM
An initial objective score below the threshold triggers one automatic correction; videos still below the threshold proceed to human review.
04 / HARNESS & ARCHITECTURE
The task coordination system (left) records execution state, artifacts and stage summaries as evidence for the experience repository (right). The repository reviews reusable rules and returns relevant experience to later Building and Rendering stages, closing the loop from execution records to experience and back to generation.
Building and Rendering Agents retrieve experience before generation. Their stage documents produce new candidates; the library compares each candidate with existing rules and evidence before deciding how the repository should change.
General rules apply across tasks; specialized rules are scoped to robots, grippers, scenes, tasks or assets. A no-op leaves the repository unchanged. Retrieved rules return to generation context, completing the experience loop.
Keep the current stage, iteration and feedback aligned as agents exchange results.
Retain project previews, execution logs, checks, video versions and user decisions.
Recover the task context and intermediate results to review or continue an interrupted workflow.
Generate synchronized original, following third-person and approximate first-person views from one scene. Edit camera paths on a shared motion timeline.
Program object relationships and motion for manipulation and locomotion tasks. Use feedback to refine the relevant part of the scene program.
Use one workflow across single arms, dual arms with grippers or dexterous hands, quadrupeds and humanoids.
Render one Proxy Video in different scene styles. Reference images and prompts vary appearance while the proxy guides motion and timing.