Veo 4 video generation

Capabilities and use cases

• Model ID: Use an available video-generation model from the Model Catalog (the examples use your-video-model-id as a placeholder).
• Text to video: Generate moving images directly from a text description.
• Image to video: Supply a reference image as the starting point for the generated video.
• Typical uses: Advertising clips, product demonstrations, social content and storyboard previews.

Request workflow

Video generation takes much longer than ordinary requests, so the API uses asynchronous tasks:

1. Submit a generation task and receive a task ID immediately.
2. Poll its status until it completes or fails.
3. When complete, download the video from the returned address.

Submit a generation task

POST https://api.yomiapi.com/v1/video/generations

Request parameters

size maps the longest side to a resolution tier: 1920 or more gives 1080p, 3840 or more gives 4K, and smaller values give 720p. Height greater than width produces 9:16 portrait; otherwise output is 16:9 landscape.

Python example

import requests

resp = requests.post(
    "https://api.yomiapi.com/v1/video/generations",
    headers={
        "Authorization": "Bearer YOUR_API_KEY",
        "Content-Type": "application/json"
    },
    json={
        "model": "your-video-model-id",
        "prompt": "A drone shot slowly gliding over a coastline at dusk, waves breaking against rocks, a cinematic look",
        "size": "1280x720",
        "seconds": "8"
    }
)
task = resp.json()
print(task["id"], task["status"])

Successful submission immediately returns a task object:

{
  "id": "task_xxxxxxxx",
  "task_id": "task_xxxxxxxx",
  "object": "video",
  "model": "your-video-model-id",
  "status": "queued",
  "progress": 0,
  "created_at": 1753600000
}

Query task status

GET https://api.yomiapi.com/v1/video/generations/{task_id}

Polling example

import time
import requests

headers = {"Authorization": "Bearer YOUR_API_KEY"}
task_id = "task_xxxxxxxx"  # ID returned by the submission endpoint

while True:
    r = requests.get(
        f"https://api.yomiapi.com/v1/video/generations/{task_id}",
        headers=headers
    )
    data = r.json()["data"]
    if data["status"] == "succeeded":
        print("Video URL:", data["url"])
        break
    if data["status"] == "failed":
        print("Generation failed:", data.get("error"))
        break
    time.sleep(10)

Download and save

After completion, url in the query response is the video address. You can also download through the content endpoint with an Authorization header:

GET https://api.yomiapi.com/v1/videos/{task_id}/content

Download and save generated results promptly to avoid losing access when the links expire.

OpenAI-compatible endpoints

The platform also offers equivalent OpenAI Videos-style endpoints for integrations following the OpenAI video API format:

• Submit: POST /v1/videos, with the same parameters as above.
• Query: GET /v1/videos/{task_id}, returning a video object whose status is queued, processing, succeeded or failed.

Billing and recommendations

• Billing: Video billing methods and per-task charges depend on the Model Catalog and actual charges under Usage.
• Prompts: Describe camera movement (push, pull, pan, track), subject action and scene lighting instead of stacking adjectives. Physical movement quality depends on how specifically you describe the action.
• Polling interval: Generation usually takes several minutes. Query about every 10 seconds to avoid excessive requests.