↑↓ Navigate enter Open esc Close
Dashboard

Build with image, video, audio, 3D, and text

One API for the entire generative stack. Hundreds of models on Runware are addressable by their AIR identifier and reachable through a single call.

Runware exposes a single endpoint for image, video, audio, 3D, and text generation. Models are addressed by their model identifier, and the same request shape works across every modality.

One shape, every modality

Create API key
→ Request
{
  "taskType": "imageInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "model": "xai:grok-imagine@image-quality",
  "positivePrompt": "A golden retriever on a park bench, cinematic lighting",
  "width": 1024,
  "height": 1024
}
{
  "taskType": "videoInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "model": "klingai:kling-video@3-4k",
  "positivePrompt": "Ocean waves crashing on a beach at sunset",
  "width": 3840,
  "height": 2160,
  "duration": 8
}
{
  "taskType": "audioInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "model": "fishaudio:s2.1@pro",
  "speech": {
    "text": "Welcome to Runware."
  }
}
{
  "taskType": "3dInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "model": "tripo:v3.1@0",
  "positivePrompt": "A vintage Polaroid camera, detailed mesh"
}
{
  "taskType": "textInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "model": "anthropic:claude@opus-4.8",
  "messages": [
    { "role": "user", "content": "Write a haiku about the ocean." }
  ]
}
← Response
{
  "taskType": "imageInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "imageUUID": "77da2d99-a6d3-44d9-b8c0-ae9fb06b6200",
  "imageURL": "https://im.runware.ai/image/os/a14d18/ws/2/ii/77da2d99-a6d3-44d9-b8c0-ae9fb06b6200.jpg",
  "cost": 0.05
}
{
  "taskType": "videoInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "videoUUID": "b7db282d-2943-4f12-992f-77df3ad3ec71",
  "videoURL": "https://vm.runware.ai/video/os/a14d18/ws/2/vi/b7db282d-2943-4f12-992f-77df3ad3ec71.mp4",
  "cost": 3.36
}
{
  "taskType": "audioInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "audioUUID": "f1e2d3c4-b5a6-7890-1234-567890abcdef",
  "audioURL": "https://am.runware.ai/audio/os/a14d18/ws/2/ai/f1e2d3c4-b5a6-7890-1234-567890abcdef.mp3",
  "cost": 0.0003
}
{
  "taskType": "3dInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "outputs": {
    "files": [
      {
        "uuid": "8c2e6d99-7404-43e6-a8f0-7a3b8d8d9f0c",
        "url": "https://im.runware.ai/image/os/a14d18/ws/5/ii/8c2e6d99-7404-43e6-a8f0-7a3b8d8d9f0c.glb"
      }
    ]
  },
  "cost": 0.3
}
{
  "taskType": "textInference",
  "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
  "text": "Ocean breathes the dawn\nWaves whisper their endless songs\nMoon hangs silent watch",
  "finishReason": "stop",
  "cost": 0.0005
}

415 hosted models. Plus thousands of community and user-uploaded models that run on any of our supported architectures.

Browse all models

Getting started

  1. Sign up for a Runware account.
  2. Generate an API key from your dashboard.
  3. Read Authentication to learn how to send the key with your requests.
  4. Browse Models to find the right one for your use case.

The fastest path is the TypeScript SDK or Python SDK . Both wrap REST and WebSocket transports and validate requests against the model's JSON Schema before sending. For AI agents, see the MCP integration .

Your first request

The same call in cURL, TypeScript, or Python.

curl -X POST https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[{
    "taskType": "imageInference",
    "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
    "model": "xai:grok-imagine@image-quality",
    "positivePrompt": "A golden retriever on a park bench, cinematic lighting",
    "width": 1024,
    "height": 1024
  }]'
import { createClient } from '@runware/sdk'

const client = createClient({ apiKey: process.env.RUNWARE_API_KEY })

const images = await client.run({
  taskType: 'imageInference',
  model: 'xai:grok-imagine@image-quality',
  positivePrompt: 'A golden retriever on a park bench, cinematic lighting',
  width: 1024,
  height: 1024,
})

console.log(images[0].imageURL)
import asyncio
from runware import Runware

async def main():
    async with Runware(api_key="YOUR_API_KEY") as client:
        images = await client.run({
            "taskType": "imageInference",
            "model": "xai:grok-imagine@image-quality",
            "positivePrompt": "A golden retriever on a park bench, cinematic lighting",
            "width": 1024,
            "height": 1024,
        })
        print(images[0]["imageURL"])

asyncio.run(main())

Core API concepts

Task-based architecture

Every Runware request is a task. Each task is processed independently, so you can send one task or batch many in a single call. Long-running tasks (video, large images) run asynchronously and deliver results as they complete.

Anatomy of a request

Hover any field below to see what it does. The same shape applies to every task type on Runware.

[
  {
    "taskType": "imageInference",
    "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
    "model": "xai:grok-imagine@image-quality",
    "positivePrompt": "A golden retriever on a park bench",
    "width": 1024,
    "height": 1024
  }
]
1
Array of tasks

Every request is an array. Send one task or batch many in a single call.

2
Task object

Each item in the array is one task. The same object shape works for image, video, audio, 3D, and text.

3
Task type

Picks the operation. imageInference, videoInference, audioInference, 3dInference, textInference, plus utilities.

4
Task UUID

A UUID v4 you generate. Used to match the response and to trace the task in your dashboard.

5
Model

Address any model on the platform by its model identifier: creator:family@version.

Common response structure

Every response echoes taskType and taskUUID from the request, and adds the modality-specific output: imageURL, videoURL, audioURL, mesh files for 3D, or text for LLMs. Set includeCost: true on the request to get a cost field back.

When a task produces multiple results, each completes independently and arrives in its own message. Output URLs are retained for 7 days by default. Set ttl on the request to change the retention.

{
  "data": [
    {
      "taskType": "imageInference",
      "taskUUID": "a770f077-f413-47de-9dac-be0b26a35da6",
      "imageUUID": "77da2d99-a6d3-44d9-b8c0-ae9fb06b6200",
      "imageURL": "https://im.runware.ai/image/os/a14d18/ws/2/ii/a770f077-f413-47de-9dac-be0b26a35da6.jpg",
      "cost": 0.0013
    },
    {
      "taskType": "videoInference",
      "taskUUID": "b880f077-e514-58ef-0ebd-ce1c37b46eb7",
      "videoUUID": "b7db282d-2943-4f12-992f-77df3ad3ec71",
      "videoURL": "https://vm.runware.ai/video/os/a14d18/ws/2/vi/b7db282d-2943-4f12-992f-77df3ad3ec71.mp4",
      "cost": 0.18
    }
  ]
}
Ask AI
Send feedback