Kling video

Kling video generation is exposed as an asynchronous Transtreams task flow. It is not a chat-completions endpoint.

Use kling-v3 plus mode for the Standard/Pro text-to-video and image-to-video routes. Use kling-v3-omni for the model displayed as Kling: Video v3.0 Omni, and call the native Omni route.

Routes

Authentication

Authorization: Bearer <ANYINT_API_KEY>

Common request fields

Image and video reference fields

Text-to-video example

Use model_name: "kling-v3" and set the edition with mode. If the model catalog shows kling-v3-std or kling-v3-pro, call the upstream route with model_name: "kling-v3" and mode: "std" or "pro".

curl -X POST "https://gateway.api.anyint.ai/transtreams/kling/v1/videos/text2video" \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "kling-v3",
    "model": "kling-v3",
    "mode": "std",
    "prompt": "A cinematic night street scene with rain, neon reflections, and slow camera movement.",
    "duration": 5,
    "aspect_ratio": "16:9",
    "generate_audio": false
  }'

Image-to-video example

Replace image with a publicly reachable HTTPS image URL from your own storage.

curl -X POST "https://gateway.api.anyint.ai/transtreams/kling/v1/videos/image2video" \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "kling-v3",
    "model": "kling-v3",
    "mode": "pro",
    "prompt": "Animate the subject with a slow cinematic dolly-in.",
    "image": "https://your-domain.com/source-image.png",
    "duration": 5,
    "generate_audio": true,
    "voice_id": "voice_demo_1"
  }'

Kling V3 Omni video example

The model catalog displays Omni as Kling: Video v3.0 Omni. Send model_name: "kling-v3-omni" and call /videos/omni-video; do not reuse the Standard/Pro /videos/text2video route for Omni.

curl -X POST "https://gateway.api.anyint.ai/transtreams/kling/v1/videos/omni-video" \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "kling-v3-omni",
    "model": "kling-v3-omni",
    "prompt": "A cinematic Tokyo street at night with neon reflections on wet pavement.",
    "mode": "std",
    "duration": "5",
    "aspect_ratio": "16:9",
    "sound": "off"
  }'

When using a reference video, include video_list.

{
  "model_name": "kling-v3-omni",
  "model": "kling-v3-omni",
  "prompt": "Preserve the reference motion and restyle it as a cyberpunk city scene.",
  "mode": "pro",
  "duration": "5",
  "sound": "off",
  "video_list": [
    { "video_url": "https://your-domain.com/reference.mp4" }
  ]
}

When using reference images with Omni video, include image_list. Do not send image_url as a top-level field, and do not use reference_image_urls for this route. Reference images from the prompt with <<<image_1>>>, <<<image_2>>>, and so on.

{
  "model_name": "kling-v3-omni",
  "model": "kling-v3-omni",
  "prompt": "Make the person in <<<image_1>>> wave toward the camera while keeping the background consistent.",
  "mode": "pro",
  "duration": "5",
  "aspect_ratio": "16:9",
  "sound": "off",
  "image_list": [
    {
      "image_url": "https://your-domain.com/person.png",
      "type": "first_frame"
    }
  ]
}

Each image_list item must include image_url, which can be a public HTTPS image URL or Base64 input. The optional type can be first_frame or end_frame; when using an end frame, also provide a first frame.

Query a task

The create response can return the task ID as task_id, id, data.task_id, or result.task_id.

Create response fields

curl -X GET "https://gateway.api.anyint.ai/transtreams/kling/v1/videos/text2video/$TASK_ID" \
  -H "Authorization: Bearer $ANYINT_API_KEY"

For image-to-video tasks, use /transtreams/kling/v1/videos/image2video/$TASK_ID.

For Omni video tasks, use /transtreams/kling/v1/videos/omni-video/$TASK_ID.

Common status fields are task_status and status. Treat succeed, succeeded, success, finished, and completed as success states. Generated video URLs commonly appear in task_result.videos[0].url.

Query response fields

Common mistakes

• Sending kling-v3-std as the upstream model_name instead of model_name: "kling-v3" plus mode: "std".
• Sending kling-v3-omni to /videos/text2video instead of /videos/omni-video.
• Sending image_url instead of image on the image-to-video route.
• Sending top-level image_url or reference_image_urls to the Omni video route instead of image_list[].image_url.
• Polling /videos/generations/{task_id} instead of the matching native route, such as /videos/text2video/{task_id}.
• Reusing OpenAI chat-completions examples for Kling video models.