kling-image-to-video

name	kling-image-to-video
description	Generate AI videos from static images using Kling's Image-to-Video API with audio-visual sync. Use this skill whenever the user wants to animate an image, convert a photo to video, create an AI video from a picture, add motion to a still image, generate video from image with camera movement or voice, or use Kling image2video — even if they don't say "Kling" explicitly.
license	MIT

Kling Image-to-Video (Audio-Visual Sync)

Generate high-quality videos from static images using Kling AI. Supports multiple model versions (kling-v1 through kling-v2-6), professional quality mode, camera control, motion brush trajectories, end-frame control, and voice synthesis. Videos can be 5s or 10s.

This is an async skill: submitting the task returns a task_id immediately; the video becomes available only after polling until task_status is succeed.

Read references/create-task-api.md for the full submit-task specification and Edge Function code. Read references/query-task-api.md for the full query-task specification and Edge Function code.

---

End-to-End Workflow

Generation-time usage (Agent direct call)

Use this pattern when working directly as an agent (no application frontend needed).

const apiKey = process.env["INTEGRATIONS_API_KEY"]!; // platform_managed — injected by platform

// Step 1: Submit the image-to-video task
async function submitImage2VideoTask(params: {
  image: string;           // Required: Base64 string or accessible URL (jpg/jpeg/png, max 10MB)
  prompt?: string;         // Optional: text description, max 2500 chars
  model_name?: string;     // Optional: default "kling-v2-6"
  mode?: "std" | "pro";   // Optional: default "pro"
  duration?: "5" | "10";  // Optional: default "5"
  image_tail?: string;     // Optional: end-frame control image
  negative_prompt?: string;
  cfg_scale?: number;      // [0,1], not supported in v2.x
  voice_list?: { voice_id: string }[];
  sound?: "on" | "off";
  static_mask?: string;    // Optional: static motion brush mask image (Base64 or URL)
  camera_control?: {
    type: "simple" | "down_back" | "forward_up" | "right_turn_forward" | "left_turn_forward";
    config?: { horizontal?: number; vertical?: number; pan?: number; tilt?: number; roll?: number; zoom?: number };
  };
  external_task_id?: string;
  callback_url?: string;
}): Promise<string> {
  const response = await fetch("https://[e-mail]/v1/videos/image2video", {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      "X-Gateway-Authorization": `Bearer ${apiKey}`,
    },
    body: JSON.stringify(params),
  });

  if (!response.ok) throw new Error(`HTTP error: ${response.status}`);
  const json = await response.json();
  if (json.code !== 0) throw new Error(`API error ${json.code}: ${json.message}`);
  return json.data.task_id;
}

// Step 2: Poll until the video is ready
async function pollUntilDone(taskId: string): Promise<{ id: string; url: string; duration: string }[]> {
  const POLL_INTERVAL_MS = 7000;
  const TIMEOUT_MS = 10 * 60 * 1000; // 10 minutes
  const deadline = Date.now() + TIMEOUT_MS;

  while (Date.now() < deadline) {
    await new Promise(r => setTimeout(r, POLL_INTERVAL_MS));

    const res = await fetch(
      `https://[e-mail]/v1/videos/image2video/${taskId}`,
      {
        method: "GET",
        headers: {
          "Content-Type": "application/json",
          "X-Gateway-Authorization": `Bearer ${apiKey}`,
        },
      }
    );
    if (!res.ok) throw new Error(`Poll HTTP error: ${res.status}`);
    const json = await res.json();
    if (json.code !== 0) throw new Error(`Poll API error ${json.code}: ${json.message}`);

    const { task_status, task_status_msg, task_result } = json.data;
    if (task_status === "succeed") return task_result.videos;
    if (task_status === "failed") throw new Error(`Task failed: ${task_status_msg}`);
    // submitted / processing — keep polling
  }
  throw new Error(`Task ${taskId} timed out after 10 minutes`);
}

// Full workflow
async function generateVideo(image: string, prompt?: string) {
  const taskId = await submitImage2VideoTask({ image, prompt });
  console.log(`Task submitted: ${taskId}`);
  const videos = await pollUntilDone(taskId);
  return videos; // [{ id, url, duration }]
}

Generation-time file download (required):

The video URL returned by the generation API is an ephemeral CDN link with a 30-day TTL. After obtaining the URL during generation (Agent direct call scenario), immediately use the Bash tool to download the file locally so the user can view the result.

curl -L -o ./output_video.mp4 "<generated video URL>"

Complete generation-time workflow (including download step):

1. Call submitImage2VideoTask to obtain the task_id
2. Call pollUntilDone to poll until status is succeed and obtain the video URL
3. Use the Bash tool to run curl -L -o <local-path>.mp4 "<url>" to download the video locally
4. Inform the user of the file path where the video has been saved

Note: The upstream CDN link expires after 30 days; download immediately after obtaining the URL.

---

Post-generation usage (in-app via Edge Function)

For production apps, use two Edge Functions — one per endpoint — so the platform API key never reaches the browser. After the query Edge Function retrieves the video URL, transfer it to Supabase Storage for persistence.

Read references/create-task-api.md for the complete submit Edge Function (edge-functions/kling-submit-image2video.ts). Read references/query-task-api.md for the complete query Edge Function (edge-functions/kling-query-image2video.ts) including Supabase Storage transfer.

Frontend polling loop:

// 1. Submit task via Edge Function
const { data: submitData, error: submitError } = await supabase.functions.invoke(
  "kling-submit-image2video",
  { body: { image, prompt, model_name: "kling-v2-6", mode: "pro", duration: "5" } }
);
if (submitError) throw submitError;
const taskId: string = submitData.data.task_id;

// 2. Poll via Edge Function until done
const POLL_INTERVAL_MS = 7000;
const TIMEOUT_MS = 10 * 60 * 1000;
const deadline = Date.now() + TIMEOUT_MS;

while (Date.now() < deadline) {
  await new Promise(r => setTimeout(r, POLL_INTERVAL_MS));

  const { data: queryData, error: queryError } = await supabase.functions.invoke(
    ,
    { : { : taskId } }
  );
   (queryError)  queryError;

   { task_status, task_status_msg, task_result } = queryData.;
   (task_status === ) {
    
     task_result.;
  }
   (task_status === )   ();
}
  ();

---

Parameter Summary

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

Parameter	Type	Required	Description
image	string	Yes	Reference image (Base64 or URL), jpg/jpeg/png, ≤10MB, ≥300px, aspect ratio 1:2.5~2.5:1
prompt	string	No	Positive text prompt, ≤2500 characters
model_name	string	No	Model version, default kling-v2-6
mode	string	No	Generation mode: std / pro, default pro
duration	string	No	Video duration (seconds): 5 or 10, default 5
image_tail	string	No	End-frame control image, same format requirements as image
static_mask	string	No	Static motion brush mask image (Base64 or URL); aspect ratio must match image
camera_control	object	No	Camera movement control; mutually exclusive with image_tail and dynamic_masks/static_mask
camera_control.type	string	No	Preset type: simple (custom) / down_back / forward_up / right_turn_forward / left_turn_forward
camera_control.config.horizontal	number	No	Left/right translation [-10, 10]; only when type=simple, only one config field may be non-zero
camera_control.config.vertical	number	No	Up/down translation [-10, 10]
camera_control.config.pan	number

---

Notes

• Key security: INTEGRATIONS_API_KEY must only be read server-side in Edge Functions; never expose it to the frontend.
• Error handling: Always handle 429 (quota exceeded) and 402 (insufficient balance).
• Billing: The submit task endpoint (api-eLMlJj3KJD89) has billing enabled — original price ¥100.80, discount price ¥84.00 (per 84 billing units). The query endpoint (api-rLobzpqX85m9) is free. Avoid re-submitting duplicate tasks to minimise unnecessary charges.
• Mutually exclusive parameters: image+image_tail, dynamic_masks/static_mask, and camera_control cannot be used simultaneously.
• Base64 format: Do not include the data:image/xxx;base64, prefix in Base64 image data; pass only the encoded string itself.
• Mask images: The aspect ratio of a mask image must match the input image; otherwise the task will fail.
• Video expiry: Generated video CDN links are automatically cleared after 30 days — download or transfer to Supabase Storage promptly.