Overview

Kling 2.5 focuses on robotics and industrial simulation. It supports physically plausible movements, constraint-aware motion planning, and exports for RGB, depth, normals, and optical flow — perfect for digital twins and training datasets.

Required Assets

Scene file: glTF (.glb) or FBX uploaded via multipart form data.
Optional motion curves: JSON describing joint positions or path splines.
Materials and HDRIs packaged inside the scene file or referenced by URL.

Example Request

curl -X POST "https://api.transendai.net/v1/videos/kling2.5/generation" \
  -H "Authorization: Bearer $TRANSEND_API_KEY" \
  -F "[email protected]" \
  -F 'payload={
    "prompt": "Robotic arm assembling a circuit board with pick-and-place motions.",
    "duration": 20,
    "physics": { "frame_rate": 120, "collision_detection": true },
    "sensors": ["rgb", "depth", "optical_flow"],
    "constraints": { "joint_speed_rad": 0.05 }
  }'

Response Payload

{
  "task_id": "task_kling_a81",
  "status": "succeeded",
  "result": {
    "streams": {
      "rgb": "https://edge.transendai.net/task_kling_a81/rgb.mp4",
      "depth": "https://edge.transendai.net/task_kling_a81/depth.tar",
      "optical_flow": "https://edge.transendai.net/task_kling_a81/flow.tar"
    },
    "physics_log": "https://edge.transendai.net/task_kling_a81/physics.json",
    "duration": 20
  }
}

Depth and optical flow streams are delivered as tar archives containing 16-bit PNG frames unless otherwise specified.

Motion Planning

Provide either high-level prompts or explicit joint trajectories:

"motion": {
  "joints": [
    { "name": "shoulder", "angles": [[0, 0.2], [0.5, 0.35], [1.0, 0.3]] },
    { "name": "elbow", "angles": [[0, 0.4], [0.5, 0.5], [1.0, 0.42]] }
  ]
}

Each array entry is [timeline_fraction, value]. Kling automatically interpolates between waypoints while respecting constraints.

Sensor Output Options

Sensor	Format	Notes
`rgb`	MP4/H.264	Color footage, default 1080p.
`depth`	PNG (16-bit)	Unit: meters.
`normals`	PNG (float-encoded)	For surface analysis.
`optical_flow`	PNG (float-encoded)	Use `decode_flow.py` helper from the SDK.

Error Codes

Code	Condition	Remediation
`400`	Scene missing or malformed JSON.	Validate multipart payload and JSON syntax.
`409`	Constraint violation detected.	Reduce joint speeds or relax motion bounds.
`415`	Unsupported asset format.	Convert models to `.glb` or `.fbx`.

Operational Guidance

Keep scene files under 200 MB; larger assets should be uploaded to S3 with signed URLs.
Use frame rates ≥ 60 FPS for robotics applications that require smooth control curves.
For custom sensors, coordinate with support to enable private plugins.

Sensor decoding utilities
Observability guide for task tracing

Kling 2.5 API