Generate Videos with Kling

What you can do

Use WorldRouter’s async video task routes to generate videos with Kuaishou Kling. Create a task on the route that matches your input, then poll its status until the video is ready, and download the output video from the task response.

Before you start

• You need a team-scoped API key with available credit. Kling bills by generated second and settles only after the task reaches a terminal status.
• Use the same base URL from Quickstart. Kling video lives under the /v1/videos surface.
• Kling request bodies use PascalCase fields, and the status route must match the route you created the task on.

Supported models

Kling accepts a Duration of 5 or 10 seconds. Send it as a string ("5"); a numeric 5 is also accepted and coerced to a string before it reaches the upstream.

Quick start flow

1. 1Create a video task on the endpoint that matches your model and input.
2. 2Save the task id returned by the create response.
3. 3Poll the matching status route until the job reaches a terminal state.
4. 4When status becomes succeeded, download the video from content.video_url.

Create a task

Kling request bodies use PascalCase fields. Pick the endpoint that matches your input:

• Text-to-video — POST /v1/videos/text2video (kling-v2-6, kling-v3).
• Image-to-video — POST /v1/videos/image2video (kling-v2-6, kling-v3) with an ImageList.
• Omni — POST /v1/videos/omni-video (kling-v3-omni), optionally with a first-frame image.

curl -X POST "https://inference-api-pre-d80ca3.worldrouter.ai/v1/videos/text2video" \
-H "Authorization: Bearer your_api_key" \
-H "Content-Type: application/json" \
-d '{
  "Model": "kling-v3",
  "Prompt": "A cinematic shot of a city street at sunset.",
  "Duration": "5",
  "Mode": "std",
  "AspectRatio": "16:9"
}'

To guide generation with a first frame, send the request to /v1/videos/image2video (or /v1/videos/omni-video for kling-v3-omni) with an ImageList. Use the image2video route, not text2video — text2video does not accept an ImageList and rejects the request. image2video derives the ratio from the image, so omit AspectRatio:

curl -X POST "https://inference-api-pre-d80ca3.worldrouter.ai/v1/videos/image2video" \
-H "Authorization: Bearer your_api_key" \
-H "Content-Type: application/json" \
-d '{
  "Model": "kling-v3",
  "Prompt": "Animate this scene with a slow dolly-in.",
  "Duration": "5",
  "Mode": "std",
  "ImageList": [
    { "ImageUrl": "https://your-host.example.com/first-frame.png", "Type": "first_frame" }
  ]
}'

Typical create response:

{
"id": "job-123",
"model": "kling-v3",
"status": "queued",
"kling_action_family": "text",
"requestId": "req-123"
}

Save the returned task id, then continue to Check task status.

Check task status

Poll until the task reaches a terminal state. The status route must match the route you created the task on:

• Text-to-video — GET /v1/videos/text2video/{task_id}
• Image-to-video — GET /v1/videos/image2video/{task_id}
• Omni — GET /v1/videos/omni-video/{task_id}

curl "https://inference-api-pre-d80ca3.worldrouter.ai/v1/videos/text2video/job-123" \
-H "Authorization: Bearer your_api_key"

Typical status response:

{
"id": "job-123",
"model": "kling-v3",
"status": "succeeded",
"kling_action_family": "text",
"content": {
  "video_url": "https://media.example.com/kling/output.mp4"
}
}

WorldRouter normalizes upstream task states into:

• queued
• running
• succeeded
• failed
• cancelled
• expired

Download the output

Only download the result after the task status becomes succeeded:

curl -L "https://media.example.com/kling/output.mp4" -o output.mp4

Read the final video URL from content.video_url in the task status response, not from the create response.

Notes and limits

• Billing is per generated second and settles when the task reaches a terminal state. Kling pricing varies by Mode (std/pro). Token-count fields are intentionally empty for these models — only the Credit charge moves.
• A task is rejected at create time if your team balance is too low.
• The number of in-flight, or pending, tasks per user is capped. Exceeding the cap returns 429.
• Task reads are owner-scoped: you can only poll a task created by the same authorized account, team, or API key. Cross-owner reads return 403 media_resource_access_denied.
• The status query must use the same capability route used to create the task.
• List routes are disabled. A GET without a task id is not supported.
• Input image URLs must be publicly reachable so the upstream model can fetch them.

Troubleshooting

See also

• Grok video guide: the Grok Imagine async video API.
• Seedance video guide: the Seedance async task API.
• Generate Images (OpenAI-compatible): image generation, including Grok Imagine image.
• API reference: chat completions for text models.
• Models: full catalog with live pricing.
• Quickstart: API key and base URL.