Video Generation (Async)

Video generation takes a long time (from tens of seconds to several minutes), so 4ALL API uses an asynchronous submit + poll protocol: first submit a task to get a task_id, then poll to query the result.

1. Submit a task

POST https://api.4allapi.com/v1/video/generations

curl https://api.4allapi.com/v1/video/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AIPROXY_KEY" \
  -d '{
    "model": "veo-3.1",
    "prompt": "A drone view gliding over the coastline at dawn, cinematic, natural light",
    "size": "1280x720",
    "seconds": "8"
  }'

Response:

{ "task_id": "video-task-xxx", "status": "submitted" }

Parameter	Description
`model`	Video model, such as `veo-3.1`, `sora-2`, `kling-v3`, `omni-flash` video model
`prompt`	Scene and camera movement description
`size`	Resolution (landscape/portrait), according to the model’s supported range
`seconds`	Duration in seconds, according to the model’s supported range
`image`	First-frame reference image for image-to-video (supported models, URL or base64)

2. Query the result

GET https://api.4allapi.com/v1/video/generations/{task_id}

curl https://api.4allapi.com/v1/video/generations/video-task-xxx \
  -H "Authorization: Bearer $AIPROXY_KEY"

status will go through submitted → processing → succeeded / failed; when successful, it returns the video download URL:

{
  "task_id": "video-task-xxx",
  "status": "succeeded",
  "video_url": "https://.../result.mp4"
}

Polling recommendations

Poll once every 5–10 seconds; extend the total timeout according to the model duration (for example, 10 minutes);
The failed status will include the failure reason; for content-policy failures, adjust the prompt and submit again;
Video links are valid for a limited time; after generation succeeds, transfer them to your own storage as soon as possible.