Does Runway have an API? Yes: how Runway Dev works

Yes, Runway has an API. Its developer platform, Runway Dev, exposes video, image and audio models over HTTP: you start a generation such as POST /v1/text_to_video, get a task id back, and poll GET /v1/tasks/{id} until the task finishes. Runway publishes its own SDKs for Node (@runwayml/sdk) and Python (runwayml).

Every Runway fact below comes from Runway's own developer docs, read on 2026-09-28 and listed under Sources. To call it you need a Runway Dev account (getting started) and a key from the Developer Portal (Runway Dev context).

How does the Runway API work?

Runway's docs call asynchronous generation “the single most important thing to internalise” (Runway Dev context):

• Base URL https://api.dev.runwayml.com, with Authorization: Bearer and your key, plus X-Runway-Version: 2024-11-06 on every request.
• The generation request returns an id. Poll GET /v1/tasks/{id}; status is PENDING, THROTTLED, RUNNING, SUCCEEDED, FAILED or CANCELLED.
• Only SUCCEEDED carries an output array. Output URLs are temporary, so download what you keep.
• DELETE /v1/tasks/{id} cancels a running task or deletes a finished one.

curl -X POST https://api.dev.runwayml.com/v1/image_to_video \
  -d '{
    "promptImage": "https://example.com/product.jpg",
    "promptText": "A timelapse on a sunny day with clouds flying by",
    "model": "gen4.5",
    "ratio": "1280:720",
    "duration": 5
  }' \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $RUNWAYML_API_SECRET" \
  -H "X-Runway-Version: 2024-11-06"

Is there a Runway API for Python?

Yes. pip install runwayml installs the Python SDK, which has MyPy-compatible type annotations and runs on Python 3.8 and up; npm install --save @runwayml/sdk installs the Node SDK, with TypeScript bindings, for Node 18 and up (SDKs). Both read the key from RUNWAYML_API_SECRET, and wait_for_task_output() (Python) or waitForTaskOutput() (Node) polls the task for you. If you poll by hand, Runway recommends an interval of 5 seconds or more, with jitter.

Which models can I call through the Runway API?

Runway's own models and other labs'. Request bodies are per model, not per endpoint: on POST /v1/text_to_video, gen4.5 takes ratio 1280:720 or 720:1280, a duration from 2 to 10 seconds and a 1,000-character prompt (Runway Dev context). The same page warns never to carry a ratio or duration from one model to another.

What is a Runway Model Router?

A Model Router is a saved configuration with a stable config id. You call /v1/generate/video, /v1/generate/image or /v1/generate/audio with that id, and the router picks an eligible model by the cost, latency or quality preference you set, under any price cap you set. The response reports which model ran and what it cost (Model Routers). Routed requests are billed at the rate of the model the router selects (pricing).

How much does the Runway API cost?

Runway Dev bills in credits, bought “for $0.01 per credit” in the developer portal; each model has its own rate, for example gen4.5 at 12 credits per second (pricing). The per-model rates are covered in Runway API pricing per second.

Can I call the same models without Runway?

Some of them. Gen-4.5 and Aleph 2.0 are what Runway's docs call its “flagship models” (models). The Seedance, Wan and Grok Imagine families it lists are also on multi-model APIs; on Sume they are seedance-2.5, seedance-2-mini, seedance-2, seedance-2-fast, wan-3.0, grok-imagine-video-1.5 on POST /v1/videos, with bare catalog ids and each job billed at the provider's list price × 1.25 (Video generation). Runway API and Seedance compares that route for Seedance.