Skip to main content

Workflow overview

Understanding the Move API workflow is essential for building effective motion capture applications. This guide explains the fundamental process from video input to motion capture data output.

The basic workflow​

The Move API workflow consists of four main steps:

Video Upload → Take Creation → Job Creation → Processing → Outputs

1. Video upload​

Upload your video file(s) to the Move API. The API supports various video formats and can handle both single-camera and multi-camera setups.

2. Take creation​

When you submit a video for processing, the API creates a "job" - a processing task that will analyze your video and extract motion capture data.

3. Job creation​

When you submit a video for processing, the API creates a "job" - a processing task that will analyze your video and extract motion capture data.

4. Processing​

The job runs in the background, using AI, biomechanics and physics models to:

  • Detect and track human subjects and skeletal keypoints
  • Infer kinematic data
  • Infer dynamics data
  • Generate 3D motion capture data outputs

5. Output download​

Once processing is complete, the job produces outputs in several format - the final motion capture data that contains the 3D skeletal animation.

Workflow types​

Single-camera workflow​

For single-camera motion capture:

  1. Upload a single video file
  2. Select a single-camera model
  3. Process and download the take

Multi-camera workflow​

For multi-camera motion capture:

  1. Upload multiple video files from different camera angles
  2. Provide camera calibration data
  3. Select a multi-camera model
  4. Process and download the take

Key concepts​

Jobs​

A job represents a processing task. Each job has:

  • Status: Pending, Processing, Completed, Failed
  • Model: The AI model used for processing
  • Input: Video files and parameters
  • Output: The resulting take

Takes​

A take is the processed motion capture data containing:

  • 3D skeletal keypoints
  • Frame-by-frame animation data
  • Metadata about the capture
  • Export formats (FBX, BVH, USDC, USDZ, GLB, Blend, C3D, JSON, CSV, video renders)

Processing time​

Processing time varies based on:

  • Video length: Longer videos take more time
  • Model complexity: s2 and m2 models are more accurate but slower
  • Number of cameras: Multi-camera setups require more processing
  • Queue position: Jobs are processed in order

Next steps​