Video Generation

The Scenario API extends beyond static image generation to support dynamic content creation, including video. These advanced functionalities often involve longer processing times and utilize a flexible “custom” endpoint, requiring a job polling mechanism to retrieve the final results. This guide will explain how to initiate video , monitor their progress, and retrieve the completed assets.

The Custom Endpoint

For specialized generation tasks like video, the Scenario API provides a versatile custom endpoint. This endpoint allows you to interact with specific models tailored for these complex outputs.

POST https://api.cloud.scenario.com/v1/generate/custom/{modelId} - API Reference

Where {modelId} is the identifier for the specific video generation model you wish to use (e.g., model_kling-v2-1 for Kling 2.1 video model). Available models ID are here: /get-started/generation/video-generation

Request Body for Custom Generation

The payload for the custom endpoint will vary depending on the modelId and the type of generation (video or 3D). However, common parameters often include:

Parameter	Type	Description
image	string	The first frame of your model.
prompt	string	A textual description to guide the generation.
steps	integer	The number of processing steps for the generation. Higher values can lead to higher quality but longer processing times.
guidanceScale	number	Controls how closely the generation follows the prompt.
duration	number	The desired duration of the video in seconds.
fps	integer	Frames per second for the video output.

To find the correct list of parameters for a specific model you can spy network requests. You can also have the parameters of all video models here: /get-started/generation/video-generation

Job Polling: Retrieving Asynchronous Results

Unlike instant image generation, video and 3D model generation are asynchronous processes. This means that when you make a request to the custom endpoint, you will receive a jobId immediately, but the actual generation will happen in the background. You then need to periodically poll a separate endpoint to check the status and retrieve the final asset.

Polling Endpoint

GET https://api.cloud.scenario.com/v1/jobs/{jobId} - API Reference

Polling Response

The response from the polling endpoint will contain the current status of your job. You should continue polling until the status field indicates success or failed.

{
  "job": {
    "jobId": "job_abc123def456",
    "status": "processing", // Can be \"queued\", \"processing\", \"success\", \"failed\", \"canceled\"
    "progress": 0.5, // Optional: progress percentage
    "metadata": {
      "assetIds": []
    }
  },
  "creativeUnitsCost": 5
}

Once the status is success, the metadata.assetIds field will contain the IDs to your generated video or 3D model assets.

Code Examples: Video Generation with VEO3

This example demonstrates initiating a video generation and then polling for the result.

Initial Request (cURL)

curl -X POST \
  -u "YOUR_API_KEY:YOUR_API_SECRET" \
  -H "Content-Type: application/json" \
  -d
  \"{"prompt":"A lone skateboarder, silhouetted against the golden burnish of a late afternoon sun, carves gracefully along a winding, empty coastal road high above the shimmering sea. Every turn of their board kicks up subtle puffs of dust and pigment from the sunbaked asphalt, while their loose shirt flutters dynamically behind them. In the distance, waves crash gently against rocky cliffs and wind-whipped wildflowers shudder in the breeze. Soft, diffused sunlight glances off chrome wheels and the skateboard’s deck, casting long, painterly shadows that stretch toward the ocean’s horizon. The camera sweeps alongside in a smooth tracking shot at low angle, occasionally veering ahead for a fleeting over-the-shoulder glimpse of endless blue water, before dropping back to capture the kinetic grace of each turn. Audio: the rhythmic rumble of urethane wheels, the hush of wind, distant gulls, and a faint, dreamy indie guitar melody. The tone is wistful, freeing, and full of sunlit nostalgia.",
  "generateAudio":true,
  "aspectRatio":"16:9",
  "duration":8
}\" \
  https://api.cloud.scenario.com/v1/generate/custom/model_veo3?projectId=yourprojectid

Python

from scenario_sdk import Scenario
import time

client = Scenario(
    api_key="YOUR_API_KEY",
    api_secret="YOUR_API_SECRET",
)

# Step 1: Initiate Video Generation
custom_video_model_id = "your_video_model_id"  # Replace with actual video model ID

print("Initiating video generation...")
response = client.generate.run_model(
    custom_video_model_id,
    body={
        "prompt": "A lone skateboarder, silhouetted against the golden burnish of a late afternoon sun, carves gracefully along a winding, empty coastal road high above the shimmering sea. Every turn of their board kicks up subtle puffs of dust and pigment from the sunbaked asphalt, while their loose shirt flutters dynamically behind them. In the distance, waves crash gently against rocky cliffs and wind-whipped wildflowers shudder in the breeze. Soft, diffused sunlight glances off chrome wheels and the skateboard’s deck, casting long, painterly shadows that stretch toward the ocean’s horizon. The camera sweeps alongside in a smooth tracking shot at low angle, occasionally veering ahead for a fleeting over-the-shoulder glimpse of endless blue water, before dropping back to capture the kinetic grace of each turn. Audio: the rhythmic rumble of urethane wheels, the hush of wind, distant gulls, and a faint, dreamy indie guitar melody. The tone is wistful, freeing, and full of sunlit nostalgia.",
        "generateAudio": True,
        "aspectRatio": "16:9",
        "duration": 8,
    },
)

job_id = response.job.job_id
print(f"Video generation job initiated. Job ID: {job_id}")

# Step 2: Poll for Job Status (or use response.job.wait() for a simpler approach)
status = "queued"
while status not in ["success", "failure", "canceled"]:
    print(f"Polling job {job_id}... Current status: {status}")
    time.sleep(3)

    poll = client.jobs.retrieve(job_id)
    status = poll.job.status
    progress = (poll.job.progress or 0) * 100
    print(f"Progress: {progress:.2f}%")

    if status == "success":
        asset_ids = poll.job.metadata.get("assetIds", [])
        print(f"Video generation completed! Asset IDs: {asset_ids}")
    elif status in ["failure", "canceled"]:
        print(f"Video generation failed or canceled: {poll.job.error}")

Node.js

import Scenario from '@scenario-labs/sdk';

const client = new Scenario({
  apiKey: 'YOUR_API_KEY',
  apiSecret: 'YOUR_API_SECRET',
});

async function generateVideo() {
  const customVideoModelId = 'your_video_model_id'; // Replace with actual video model ID

  console.log('Initiating video generation...');

  // Step 1: Initiate Video Generation
  const response = await client.generate.runModel(customVideoModelId, {
    body: {
      prompt: 'a futuristic city with flying cars, cinematic, 4k',
      duration: 5,
      fps: 24,
    },
  });

  const jobId = response.job.jobId;
  console.log(`Video generation job initiated. Job ID: ${jobId}`);

  // Step 2: Wait for completion using the built-in .wait() helper
  const completed = await response.job.wait();

  if (completed.status === 'success') {
    const assetIds = completed.metadata?.assetIds || [];
    console.log('Video generation completed! Asset IDs:', assetIds);
  } else {
    console.error(`Video generation ended with status: ${completed.status}`);
  }
}

generateVideo();

---

👉 Download OpenAPI spec