Quickstart

Use one API to generate images, video and audio with the models available in Fusara. Configure a request in the playground, or start with the example below.

Base URL: https://fusara.ai/api/v3 · JSON requests and responses · API-key authentication

Open your Fusara profile, create an API key and keep it on your server. API generations use the same account and balance as Fusara.

Set the key in your terminal. Replace the placeholder with your own key:

export FUSARA_API_KEY='your-api-key'

JavaScript examples run on your server with Node.js. Python examples use requests: install it with python -m pip install requests. You do not need a Fusara SDK.

Send a model ID and the inputs it needs. This example uses Qwen Image 2.1 to create an image. Optional model settings use the model's published defaults.

curl 'https://fusara.ai/api/v3/run' \
  -H "Authorization: Bearer $FUSARA_API_KEY" \
  -H 'Content-Type: application/json' \
  --data-raw '{
  "model": "qwen-image21-create",
  "prompt": "A white ceramic lamp on a pale lavender background, soft studio light."
}'

To choose a different model or adjust its settings, use the model playground. Its API code tab exports the request you configured. Each model's discovery response also includes its exact request_schema.

A 202 Accepted response means the run exists and generation is still in progress. Save its numeric id; use it to retrieve the result or match a webhook.

{
  "id": 123456,
  "status": "queued",
  "cost": {
    "unit": "common_fuz",
    "expected": "40.0000",
    "status": "pending"
  }
}

The Location response header also contains the run's API URL. A submission can return 200 if the run is already terminal, including an idempotent replay. Always inspect status.

Prices, IDs and result URLs in these examples are illustrative. Use the values returned by your own request. Estimate the current price before generating if you need it.

You can receive a webhook or check the run yourself. Both return the complete run, including usable output URLs.

Replace 123456 with the ID from your submission:

curl 'https://fusara.ai/api/v3/run/123456' \
  -H "Authorization: Bearer $FUSARA_API_KEY"

While status is queued or running, wait a few seconds before checking again. Stop on succeeded, failed or canceled. A 200 response means the read succeeded; it does not mean the generation succeeded.

{
  "id": 123456,
  "status": "succeeded",
  "cost": {
    "unit": "common_fuz",
    "expected": "40.0000",
    "status": "settled",
    "charged": "40.0000"
  },
  "outputs": [
    {
      "type": "image",
      "media_id": 901,
      "url": "https://media.example/result.png",
      "url_expires_at": "2026-10-25T12:00:00Z"
    }
  ]
}

Open or download outputs[0].url. Save files you need before url_expires_at. Other models can return video, audio, text, JSON or boolean outputs; inspect each output's type.

Add a public HTTPS webhook_url when creating the run:

curl 'https://fusara.ai/api/v3/run' \
  -H "Authorization: Bearer $FUSARA_API_KEY" \
  -H 'Content-Type: application/json' \
  --data-raw '{
  "model": "qwen-image21-create",
  "prompt": "A white ceramic lamp on a pale lavender background, soft studio light.",
  "webhook_url": "https://your-app.example/webhooks/fusara"
}'

Your endpoint receives the complete result above. Save it and respond with 2xx. No webhook registration, signature-verification setup or additional result GET is required. Fusara signs every delivery; verifying that signature is optional and authenticates the sender. Without verification, sender authenticity is unverified.

See receiving webhooks for failure payloads, duplicate handling and delivery retries.

A request can be accepted and fail later. A terminal failure appears in the run or webhook, rather than as an HTTP error from the status read:

{
  "id": 123456,
  "status": "failed",
  "cost": {
    "unit": "common_fuz",
    "expected": "40.0000",
    "status": "settled",
    "charged": "0.0000"
  },
  "error": {
    "code": "execution_failed",
    "message": "The run could not be completed."
  }
}

Read error.code and error.message; preserve any usable outputs if present. For a rejected submission or network problem, follow errors and recovery. Keep your run ID even if your application stops waiting.

• Runs & results: polling, history, cancellation and safe submission retries.
• Media & files: image, video and audio inputs; reusing results.
• Balance & pricing: estimates and final charges.
• API reference: exact methods, fields and responses.
• MCP: connect an assistant to the same models and operations.