Kling Video

Kling video models are reached through Kling's native video API, proxied under /kling. Authenticate with your sk- key as usual (Authorization: Bearer sk-...); request and response bodies are Kling's own format — see Kling's official API documentation for the authoritative field list.

Like all video generation, these are asynchronous: submit a task, then poll its status.

Endpoints

Method	Path	Purpose
POST	/kling/v1/videos/text2video	Text-to-video — create a task
POST	/kling/v1/videos/image2video	Image-to-video — create a task
GET	/kling/v1/videos/text2video/{task_id}	Text-to-video — task status
GET	/kling/v1/videos/image2video/{task_id}	Image-to-video — task status

Full URL example: https://router.accels.tech/kling/v1/videos/text2video.

Example

# Create a text-to-video task (body follows Kling's native API)
curl https://router.accels.tech/kling/v1/videos/text2video \
  -H "Authorization: Bearer $ACCELSROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model_name": "kling-v1",
    "prompt": "A red kite over a green valley at sunrise",
    "duration": "5",
    "aspect_ratio": "16:9"
  }'

# Poll (task_id from the create response)
curl https://router.accels.tech/kling/v1/videos/text2video/{task_id} \
  -H "Authorization: Bearer $ACCELSROUTER_API_KEY"

The exact request fields (e.g. model_name, prompt, negative_prompt, cfg_scale, mode, duration, aspect_ratio, image for image-to-video) and the response shape are defined by Kling — refer to Kling's API docs. AccelsRouter only handles auth, routing and billing.

See also the unified Video Generation endpoint, which covers many video models through one OpenAI-style request.