Kling V3 Image-to-Video API

Overview

Kling V3 Image-to-Video is the image-to-video interface of Kuaishou's third-generation Kling video generation model. It takes a single input image and a text prompt as core inputs, switches between Standard (720p), Pro (1080p), and 4K resolutions via the mode field, and supports 4–15 second video generation.

Compared to the same-series Kling V3 Omni, V3 has a simpler API structure — the request body is flat (image / prompt / mode / duration), with no reference video, multi-shot composition, or audio generation, making it suitable for scenarios that require quickly generating dynamic video from a static image. On iCreat it is called via the asynchronous task queue kwaivgi/kling-v3/image-to-video.

kwaivgi/kling-v3/image-to-video

Core Capabilities

Image-driven video generation. Provide a single input image as a visual reference or initial frame, along with a text prompt describing the desired motion and changes, to generate dynamic video.

Three resolution tiers. Switch via the mode field: std (720p) for lowest cost, pro (1080p) for enhanced detail, 4k (2160p) for high-end delivery.

4–15 second duration. Supports video generation from a minimum of 4 seconds to a maximum of 15 seconds, billed per second — suitable for short videos, social media content, and rapid prototyping.

Chinese & English prompts. Officially supports Chinese and English prompts; the model has good comprehension of scene descriptions and action instructions in both languages.

Flat request structure. Unlike Wan 2.7 / HappyHorse 1.1 / Seedance 2.5's input + parameters nested structure, Kling V3's request body is flat — image, prompt, mode, and duration are placed directly at the JSON root level.

Model Comparison

Kling V3 Image-to-Video vs Kling V3 Omni vs Seedance 2.5 Image-to-Video

Input

Kling V3 Image-to-Video accepts a single input image and a text prompt. It does not support reference videos or character references.

Parameters

Mode mapping

Note: Kling V3 uses std (abbreviated), while the same-series Kling V3 Omni uses standard (full word). Both pro and 4k values are identical across the two.

Pricing

Base prices without reference video (this interface is image-to-video only and does not support reference video).

Total cost = unit price × output video duration

Quick Start

1. Authentication

Get your API Key from https://icreat.ai/hub/keys.

2. Submit a Task

curl --fail-with-body --connect-timeout 10 --max-time 60 \
  -X POST "https://api.icreat.ai/v1/task/submit/kwaivgi/kling-v3/image-to-video" \
  -H "Authorization: Bearer ${ICREAT_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "image": "https://example.com/example.png",
    "prompt": "Example multi-line\ntext",
    "mode": "std",
    "duration": 4
  }'

Response example:

{
  "task_id": "task-xxx"
}

3. Poll Task Status

curl --connect-timeout 10 --max-time 60 \
  -X POST "https://api.icreat.ai/v1/task/query-status" \
  -H "Authorization: Bearer ${ICREAT_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "task_id": "task-xxx"
  }'

When the task status is SUCCEEDED, you can retrieve the result.

4. Get the Result

curl --connect-timeout 10 --max-time 60 \
  -X POST "https://api.icreat.ai/v1/task/get-result" \
  -H "Authorization: Bearer ${ICREAT_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "task_id": "task-xxx"
  }'

The server will return the task output.

Output

submit → task_id → query-status → get-result

After submitting a request you receive a task_id. Poll query-status until the status becomes SUCCEEDED, then call get-result with the same task_id to retrieve the generated video.

Input Examples

Generate a 720p video

{
  "image": "https://example.com/example.png",
  "prompt": "让小狗在草地上奔跑起来",
  "mode": "std",
  "duration": 5
}

Generate a 1080p video

{
  "image": "https://example.com/example.png",
  "prompt": "让小狗在草地上奔跑起来",
  "mode": "pro",
  "duration": 5
}

Generate a 4K video

{
  "image": "https://example.com/example.png",
  "prompt": "让小狗在草地上奔跑起来",
  "mode": "4k",
  "duration": 5
}

Use Cases

Image animation & dynamization. Transform static product images, portrait photos, or concept art into dynamic video — ideal for e-commerce showcases, social media content, and creative marketing.

Rapid prototyping & concept validation. Use the lowest-cost std mode to quickly validate image+prompt generation results, then switch to pro or 4k for high-quality final output once motion direction and composition are confirmed.

Short video & social media. 4–15 second durations cover short-video needs on TikTok, Kuaishou, Xiaohongshu, and similar platforms; image+text input lowers the barrier for material preparation.

Advertising & marketing. pro mode generates 1080p ad creative clips; 4k mode is used for brand films and high-end delivery.

Education & demo videos. Use an instructional image as the initial frame, paired with natural-language action descriptions, to generate dynamic teaching demonstrations.

Limitations

Maximum duration is 15 seconds — long videos cannot be generated through this interface (use Kling V3 Omni for up to 120 seconds).

Does not support reference video, character reference, audio generation, negative prompt, aspect ratio selection, or watermark control (use Kling V3 Omni or Seedance 2.5 for these features).

The mode value is std (not standard), which differs from Kling V3 Omni's standard — note this when migrating code.

The input image must be a publicly accessible URL; local file upload is not supported. JPEG and PNG formats are recommended.

Non-Chinese/English prompts are untested and may reduce generation quality.

The same image and prompt produce different variants on each run (by design); use more detailed prompts to constrain the model for deterministic output.

FAQ

What's the difference between Kling V3 Image-to-Video and Kling V3 Omni? Which should I choose?

V3 Image-to-Video accepts a single image plus a text prompt, supports up to 15 seconds, and has a simple API structure. V3 Omni is the full-featured version supporting image/video/character reference, multi-shot composition, audio generation, and up to 120 seconds. Choose V3 when you need to quickly generate short video from a single image; choose V3 Omni for multi-modal reference, long video, or complex narrative. Both share the same 720P / 1080P / 4K base unit prices.

Are std and standard the same mode?

No. Kling V3 uses std, while Kling V3 Omni uses standard. Both map to 720p output, but the string values differ. You must update the mode value when migrating code.

What are the requirements for the image URL?

It must be a publicly accessible HTTP or HTTPS link that the server can download. Common formats such as JPEG and PNG are supported. The URL should not contain special characters or require authentication headers.

Is 4K mode worth it?

The 4K unit price of $0.42/s is 5× the cost of `std`; a 15-second video costs $6.30. Recommended only when final delivery requires 4K resolution; use std for testing and preview.

Will I be charged for failed generations?

No. Failed tasks (FAILED) incur no charges and can be safely resubmitted. Regenerating after SUCCEEDED is a new request and will be billed again.

How long can the prompt be?

Chinese must not exceed 2000 characters; English must not exceed 2000 words. \n line breaks are supported for multi-line prompts to organize scene descriptions and action instructions.