Developers

Query worlds like a database.

One endpoint. Ten modalities. Search by scene, metadata, or signal type - stream frames directly into your training pipeline with a single API key.

Read docsAPI · operational

// app.originlab.ai · web UI · same data as the API

01/Dashboard
Browse, filter,
export.

Everything the API exposes is also available through our data platform - browse catalogs, filter by metadata, build collections, and track exports.

app.originlab.ai/library

Catalogs

Games 50+

3D Environments growing

Modalities

RGB Video *

Depth Maps *

Surface Normals *

Camera Telemetry *

UrbanNightRainInteriorCrowdVehicles

Tactical FPS · 014

Open-world · 027

Urban scan · 003

Survival · 061

// POST /v1/metadata/search · stream → Zarr | Parquet

02/API
One endpoint.
Every signal.

Search, filter, and stream datasets programmatically. All responses include schema version, provenance metadata, and rights status.

Search. Stream. Ship.

The Origin Lab SDK wraps our REST API with typed models, automatic pagination, and streaming support. Query by any metadata dimension, filter across catalogs, and pipe results directly into your training loop.

Typed search filters

20+ metadata categories as structured query params. No free-text guessing.

Streaming ingest

Zarr and Parquet frames streamed on-demand. No intermediate downloads.

Versioned exports

Pin a dataset version. Re-export anytime with full provenance chain.

Provenance in every response

License ID, rights scope, and source metadata in every API response.

pipeline.py

PYTHON

import originlab

client = originlab.Client(api_key="sk-...")

scenes = client.search(filters={
    "environment.category.eq": "urban_city",
    "weather.condition.eq": "rain",
    "modalities.includes": ["depth_16", "camera_pose"],
}, limit=50)

for scene in scenes:
    frames = client.stream(scene.id, modalities=["rgb", "depth_16"], format="zarr")
    pipeline.ingest(frames)

// curl · cursor pagination · schema-versioned responses

03/API examples
cURL in, JSON out.

Query, filter, and export datasets programmatically. All responses include schema version and metadata.

search_request.sh

cURL

curl -X POST https://app.originlab.ai/api/v1/metadata/search \
  -H "X-API-Key: $ORIGIN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "filters": {
      "environment.category.eq": "urban_city",
      "weather.weather_condition.eq": "clear",
      "people.crowd_level.in": ["large_group", "crowd"]
    },
    "limit": 10
  }'

search_response.json

JSON

{
  "results": [
    {
      "id": 18842,
      "scene_id": "scn_4821_014",
      "s3_key": "samples/4821/scene-014.mp4",
      "environment": { "category": "urban_city", "setting_era": "contemporary" },
      "weather":     { "condition": "clear", "visibility": "good" },
      "people":      { "crowd_level": "large_group" }
    }
  ],
  "total": 2847,
  "next_cursor": "eyJpZCI6MTg4NDJ9"
}

Export

export_request.sh

cURL

curl -X POST https://app.originlab.ai/api/v1/export \
  -H "X-API-Key: $ORIGIN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "collection_id": 42,
    "format": "csv"
  }'

export_response.json

JSON

{
  "items": [
    {
      "id": 4821,
      "category": "tactical_fps",
      "content_type": "videogame",
      "license_type": "scoped",
      "nb_samples": 142,
      "updated_at": "2026-02-20T14:22:00Z"
    }
  ],
  "total": 312,
  "next_cursor": "eyJpZCI6NDgyMX0="
}

// schema · 20 fields · structured, queryable, composable

04/Metadata
20 queryable
categories.

20 structured categories extracted from every clip. Filter and search across any combination.

people

crowd level, individual descriptions, activities, interactions

animals

species, behavior, role, size

flora

vegetation type, density, seasonal indicators

transportation

vehicle types, motion state, camera relationship

environment

setting category, era, location type, mood

weather

conditions, visibility, intensity, atmospheric effects

time_of_day

period, light source, sky condition

lighting_color

type, key ratio, color palette, color grade, temperature

perspective

shot type, camera angle, movement, lens characteristics

focus_tracking

depth of field, tracking, transitions

occlusion

detection, type, visibility impact

visual_style

composition, framing, texture, editing pace

text_graphics

text presence, types, coverage

conversation

emotional tone, interaction type, power dynamics

physical

interaction type, intensity, contact

continuity

transitions, temporal relationships, location changes

trademarks

brand detection, product placement context

audio_design

ambient sound, dialogue processing, effects

music_score

genre, intensity, role, presence

language

spoken/subtitle languages, accents

// stream · msgpack · 60 Hz engine + 50 Hz input, synchronized

05/Capture data
Raw signals,
structured streams.

Raw input and engine telemetry captured alongside every session. Synchronized, timestamped, and delivered as structured streams.

input_capture.json

INPUT

{
  "header": {
    "version": "TLS-0.005",
    "format":  "msgpack-stream",
    "config":  { "mouse_sample_hz": 50, "key_release": true, "scroll": true },
    "obs":     { "width": 1920, "height": 1080, "fps": 30 }
  },
  "events": [
    { "t": 0.00, "type": "active_window", "category": "tactical_fps" },
    { "t": 0.02, "type": "mouse_move",    "x": 966, "y": 542 },
    { "t": 0.67, "type": "key_down",      "key": "w" },
    { "t": 1.24, "type": "key_down",      "key": "space" }
  ]
}

engine_telemetry.json

TELEMETRY

{
  "header": {
    "version": "TLE-0.001",
    "config":  { "capture_hz": 60 },
    "engine":  { "type": "unreal", "version": "4.23", "renderer": "d3d12" },
    "title":   { "category": "tactical_fps", "build_id": "build_a14" },
    "camera":  { "coord_system": "unreal", "pos_unit": "cm", "fov_unit": "deg" }
  },
  "events": [
    { "t": 0.0000, "frame": 0, "px": 24510.0, "py": -8835.0, "pz": 1202.7, "yaw": 127.7, "pitch": -3.0, "fov": 90.0 },
    { "t": 0.0167, "frame": 1, "px": 24509.9, "py": -8834.9, "pz": 1202.0, "yaw": 127.6, "pitch": -2.6, "fov": 90.0 }
  ]
}

// X-API-Key · 60 req/min · stable error codes · request_id on every response

06/Quickstart
Auth, SDKs,
limits.

Everything you need to make a first call. Bearer-style API key, predictable error shape, generous limits.

Authentication

Pass X-API-Key on every request. Keys are environment-scoped (sandbox / prod) and rotate without downtime.

SDKs

PythonTypeScriptGoRust · soon

Rate limits

60 req/min on search, 10 concurrent stream sessions, 5 active exports. Lifted on request for production workloads.

Errors

Stable error codes, machine-readable retry hints, and a request_id on every response for support triage.

error_response.json

JSON

{
  "error": {
    "code": "rate_limit_exceeded",
    "message": "60 requests/min exceeded - retry after 12s.",
    "retry_after": 12,
    "request_id": "req_01HZX9F2M4P"
  }
}

Start building today.

Get your API key, explore the docs, and stream your first dataset in minutes. Same-day access for qualified teams.

Read the docs

Query worlds like a database.

01/DashboardBrowse, filter,export.

02/APIOne endpoint.Every signal.

Search. Stream. Ship.

Typed search filters

Streaming ingest

Versioned exports

Provenance in every response

03/API examplescURL in, JSON out.

04/Metadata20 queryablecategories.

05/Capture dataRaw signals,structured streams.

06/QuickstartAuth, SDKs,limits.

Authentication

SDKs

Rate limits

Errors

Start building today.

01/Dashboard
Browse, filter,
export.

02/API
One endpoint.
Every signal.

03/API examples
cURL in, JSON out.

04/Metadata
20 queryable
categories.

05/Capture data
Raw signals,
structured streams.

06/Quickstart
Auth, SDKs,
limits.