kling-text-to-video

Generate short videos from text prompts using the Kling AI text-to-video API. Use this skill whenever the user wants to create a video from a script, scene description, or keywords — for e-commerce ma

SKILL.md

GitHub

Kling Text-to-Video

Overview

Converts text descriptions into short videos using the Kling AI model, supporting asynchronous task polling (up to 10 minutes). Suitable for e-commerce marketing, creative promotion, educational content, and entertainment production.

Property Value
Submit Endpoint POST https://app-dohnbj5by5mp-api-qYGWo8XA7JVY.gateway.appmedo.com/v1/videos/text2video
Query Endpoint GET https://app-dohnbj5by5mp-api-oLpZ7eD5j2Pa.gateway.appmedo.com/v1/videos/text2video/{id}
Auth Bearer Token (platform_managed)
Request Format application/json
Response Format JSON (includes video URL)
Note The submit endpoint is used to create the task and is billed; the query endpoint is used to poll for results and is free

Main Request Parameters (Submit)

Parameter Type Required Description
prompt string Yes Video generation prompt
model_name string No Model name: kling-v1, kling-v1-6, kling-v2-master, kling-v2-1-master, kling-v2-5-turbo
negative_prompt string No Negative prompt
aspect_ratio string No Video aspect ratio: 16:9, 9:16, 1:1
duration string No Video duration in seconds: 5 or 10

For complete parameter descriptions and code, see references/submit-api.md and references/query-api.md.


End-to-End Async Workflow

Kling video generation is an async task: first submit the task to obtain a task_id, then poll the query endpoint until the status becomes succeed or failed.

// Full async workflow: submit → poll → retrieve video URL
async function generateAndWait(prompt: string, options?: {
  model_name?: string;
  aspect_ratio?: string;
  duration?: string;
  negative_prompt?: string;
}) {
  // Step 1: Submit task
  const { taskId } = await submitTextToVideo(prompt, options);

  // Step 2: Poll until complete
  const POLL_INTERVAL_MS = 7000;          // Recommended 5–10 seconds
  const TIMEOUT_MS = 10 * 60 * 1000;     // Max 10 minutes
  const deadline = Date.now() + TIMEOUT_MS;

  while (Date.now() < deadline) {
    await new Promise(r => setTimeout(r, POLL_INTERVAL_MS));
    const result = await queryTask(taskId);
    if (result.task_status === "succeed") return result;
    if (result.task_status === "failed")  throw new Error(`Task failed: ${result.task_status_msg}`);
    // submitted / processing → continue polling
  }
  throw new Error(`Task ${taskId} timed out after 10 minutes`);
}

Task status descriptions:

  • submitted — Task submitted, waiting to be processed
  • processing — Task is being processed
  • succeed — Task completed successfully; task_result.videos contains the video URL
  • failed — Task failed

Generation-Time Usage (Direct Agent Call)

Use the built-in scripts for generation-time calls. The scripts read INTEGRATIONS_API_KEY from the environment.

The Bash tool timeout MUST be set to 600000ms (600 seconds).

Submit + poll (all-in-one):

python3 <skill-path>/scripts/generate_text_to_video.py \
  --prompt "A serene mountain stream flowing through a forest" \
  --aspect-ratio 16:9 \
  --duration 5 \
  --mode std \
  --output-dir /tmp/text2video

Resume polling an existing task:

python3 <skill-path>/scripts/query_text_to_video.py --task-id "<task_id>" --output-dir /tmp/text2video

The scripts print one JSON line:

  • On success: {"status":"succeed","task_id":"...","videos":[{"url":"...","file":"..."}]}
  • If still processing: {"status":"processing","task_id":"..."}

On failure they print an error to stderr and exit with a non-zero code.

Generation-time file download (required):

Video URLs are ephemeral CDN links. If --output-dir is not passed, download immediately:

curl -L -o /tmp/generated_video.mp4 "<task_result.videos[0].url>"

For full parameter details, see references/submit-api.md (submit) and references/query-api.md (query).


Post-Generation Usage (In-App via Edge Function)

In the application, proxy requests through two separate Edge Functions: one for submitting tasks and one for querying status. The query Edge Function transfers the video URL to Supabase Storage upon task completion and returns a persistent public URL to the frontend.

For the complete Edge Function code and frontend call patterns see the "Post-Generation Usage" sections in references/submit-api.md and references/query-api.md.