> ## Documentation Index
> Fetch the complete documentation index at: https://docs.qingbo.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Kling Series

> Kuaishou Kling video generation — v2.6 / v3 / v3-omni / video-o1, four generations covering text-to-video, image-to-video, first/last frame, multimodal reference, and reasoning-enhanced

Kuaishou Kling video generation series.

**Four-generation lineup**:

* **kling-v2.6** — Classic stable release, 720P silent / 1080P with optional audio, further refined semantic adherence and motion stability
* **kling-v3** — Next-generation base model, adds **4K resolution**, T2V duration extended to **15 seconds**, native audio support
* **kling-v3-omni** — Unified **multimodal interface** for the v3 family — T2V, I2V, and video reference all share one endpoint, with `<<<image_N>>>` reference syntax in the prompt
* **kling-video-o1** — First **reasoning-enhanced** video model, performs deep planning over the prompt and reference assets before generation, delivering best-in-class physical consistency, complex motion, and long-form motion semantic adherence

Billed by **resolution**, in `$/second`, with `resolution` selecting 720p / 1080p / 4K.

## Pricing

| Model            | 720P      | 1080P      | 4K          | Notes                                                                                          |
| ---------------- | --------- | ---------- | ----------- | ---------------------------------------------------------------------------------------------- |
| `kling-v2.6`     | `$0.0391` | `$0.06641` | —           | 1080P with audio uplifts to `$0.159375`                                                        |
| `kling-v3`       | `$0.0714` | `$0.0952`  | `$0.455345` | With audio: 720P `$0.1071` / 1080P `$0.1428` / 4K same price                                   |
| `kling-v3-omni`  | `$0.0714` | `$0.0952`  | `$0.455345` | With audio: 720P `$0.0952` / 1080P `$0.119`; video reference: 720P `$0.1071` / 1080P `$0.1428` |
| `kling-video-o1` | `$0.0714` | `$0.0952`  | —           | Video reference: 720P `$0.1071` / 1080P `$0.1428`                                              |

<Note>
  Prices are **per second**; actual charge = unit price × `duration`. Enabling `audio` or attaching a `video_list` video reference switches to the corresponding higher tier.
</Note>

## Examples

<CodeGroup>
  ```bash Text-to-video (minimal) theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "kling-v3",
      "prompt": "A Shiba Inu in a spacesuit walking on the moon, cinematic lighting",
      "duration": 5,
      "resolution": "1080p",
      "aspect_ratio": "16:9",
      "audio": true
    }'
  ```

  ```bash Image-to-video (first frame) theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "kling-v2.6",
      "action": "image2video",
      "prompt": "The character walks slowly toward the camera, hair gently moved by the breeze",
      "first_frame_image": "https://cdn.example.com/portrait.jpg",
      "duration": 5,
      "resolution": "1080p"
    }'
  ```

  ```bash Omni multimodal reference theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "kling-v3-omni",
      "action": "reference",
      "prompt": "Place the character from <<<image_1>>> running through the forest scene from <<<image_2>>>",
      "image_urls": [
        "https://cdn.example.com/char.jpg",
        "https://cdn.example.com/forest.jpg"
      ],
      "duration": 6,
      "resolution": "1080p",
      "aspect_ratio": "16:9"
    }'
  ```

  ```bash O1 reasoning-enhanced (video reference) theme={"system"}
  curl -X POST https://www.qingbo.dev/v1/tasks \
    -H "Authorization: Bearer $WAVE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "kling-video-o1",
      "action": "reference_video",
      "prompt": "Keep the character look from <<<image_1>>>, follow the camera movement of the reference video",
      "image_urls": ["https://cdn.example.com/char.jpg"],
      "video_list": [
        {
          "video_url": "https://cdn.example.com/ref.mp4",
          "refer_type": "feature",
          "keep_original_sound": "no"
        }
      ],
      "duration": 10,
      "resolution": "1080p"
    }'
  ```
</CodeGroup>

<ResponseExample>
  ```json Submission accepted theme={"system"}
  {
    "task_id": "task-wave1775285160b950328499",
    "model": "kling-v3",
    "action": "generate",
    "status": "queued",
    "created_at": 1775285160040,
    "progress": 0
  }
  ```
</ResponseExample>

After submission, poll status with [`GET /v1/tasks/{task_id}`](/en/api-reference/task/status); see [Task System](/en/docs/task-system) for details.

## Available Models

| Model ID         | Resolution        | Duration   | Supported actions                                                                 | Highlights                                    |
| ---------------- | ----------------- | ---------- | --------------------------------------------------------------------------------- | --------------------------------------------- |
| `kling-video-o1` | 720P / 1080P      | 5 / 10 sec | `generate` · `image2video` · `first_last_frame` · `reference` · `reference_video` | Reasoning-enhanced, best physical consistency |
| `kling-v3-omni`  | 720P / 1080P / 4K | 3-15 sec   | `generate` · `image2video` · `reference` · `reference_video`                      | Unified multimodal endpoint                   |
| `kling-v3`       | 720P / 1080P / 4K | 3-15 sec   | `generate` · `image2video` · `first_last_frame`                                   | 4K + native audio                             |
| `kling-v2.6`     | 720P / 1080P      | 5 / 10 sec | `generate` · `image2video` · `first_last_frame`                                   | 1080P with optional audio, top stability      |

## Common Parameters

<ParamField body="model" type="string" required>
  Model ID; see the [Available Models](#available-models) table
</ParamField>

<ParamField body="action" type="string" default="generate">
  Operation type. Allowed values:

  * `generate` — text-to-video (T2V)
  * `image2video` — image-to-video; pair with `first_frame_image` or `image_urls`
  * `first_last_frame` — first/last frame interpolation; requires `first_frame_image` + `last_frame_image` (v2.6 / v3 / o1)
  * `reference` — multimodal reference-to-video; pair with `image_urls` (omni / o1)
  * `reference_video` — video reference-to-video; pair with `video_list` (omni / o1)
</ParamField>

<ParamField body="prompt" type="string" required>
  Video description text. In Omni / O1 you can use the `<<<image_N>>>` syntax to reference the N-th image in `image_urls` (N starts at 1)
</ParamField>

<ParamField body="aspect_ratio" type="string" default="16:9">
  Frame aspect ratio, **applies to T2V only**; image / video reference modes follow the source asset's ratio. Allowed values:

  * `16:9` — landscape widescreen
  * `9:16` — portrait
  * `1:1` — square
</ParamField>

<ParamField body="resolution" type="string" default="720p">
  Output resolution. See [Available Models](#available-models) for per-model support:

  * `720p`
  * `1080p`
  * `4K` (v3 / omni only)
</ParamField>

<ParamField body="duration" type="integer" default="5">
  Video duration in seconds. v2.6 / o1 accept only `5` or `10`; v3 / omni accept any integer from 3-15
</ParamField>

<ParamField body="image_urls" type="string[]">
  Reference image URL array. A **single-element array** triggers I2V; multiple images feed the multimodal reference for Omni / O1 (referenceable via `<<<image_N>>>`)
</ParamField>

<ParamField body="first_frame_image" type="string">
  First-frame image URL. Used with `image2video` or `first_last_frame` action (v2.6 / v3 / o1)
</ParamField>

<ParamField body="last_frame_image" type="string">
  Last-frame image URL, used with `first_last_frame` action
</ParamField>

<ParamField body="video_urls" type="string[]">
  Reference video URL array (single element). Simplified form for Omni / O1, equivalent to attaching a `video_list` entry with `refer_type=feature`
</ParamField>

<ParamField body="callback_url" type="string">
  Webhook callback URL, invoked when the task reaches a terminal state. See [Callback Mechanism](/en/docs/sync-async#async-task-webhook)
</ParamField>

<ParamField body="callback_events" type="string[]">
  Event types to push; defaults to all terminal events
</ParamField>

## Model-specific Parameters

<Tabs>
  <Tab title="kling-video-o1">
    <ParamField body="video_list" type="object[]">
      Reference video list, **at most 1 clip**. Same structure as Omni:

      * `video_url` — video URL
      * `refer_type` — `base` / `feature`
      * `keep_original_sound` — `yes` / `no`

      O1 does not support the `audio` field; for audio output use v3 / omni.
    </ParamField>
  </Tab>

  <Tab title="kling-v3-omni">
    <ParamField body="audio" type="boolean" default="false">
      Whether to generate audio. **Mutually exclusive with `video_list`** — cannot be enabled when a video reference is attached
    </ParamField>

    <ParamField body="watermark" type="boolean" default="false">
      Whether to add a watermark
    </ParamField>

    <ParamField body="video_list" type="object[]">
      Reference video list, **at most 1 clip**. Each element:

      * `video_url` — public video URL
      * `refer_type` — reference type, `base` (overall replication) / `feature` (motion / style features)
      * `keep_original_sound` — keep original audio track, `yes` / `no`

      Once `video_list` is attached, billing switches to the "video reference" tier and `audio` cannot be enabled.
    </ParamField>
  </Tab>

  <Tab title="kling-v3">
    <ParamField body="audio" type="boolean" default="false">
      Whether to generate audio. Native audio is supported across all resolutions
    </ParamField>

    <ParamField body="watermark" type="boolean" default="false">
      Whether to add a watermark
    </ParamField>
  </Tab>

  <Tab title="kling-v2.6">
    <ParamField body="audio" type="boolean" default="false">
      Whether to auto-generate audio. **Only effective at `resolution=1080p`**; **mutually exclusive with last frame** (cannot enable with `first_last_frame`)
    </ParamField>

    <ParamField body="watermark" type="boolean" default="false">
      Whether to add a watermark
    </ParamField>
  </Tab>
</Tabs>

## Special Syntax

**`<<<image_N>>>` prompt image reference** (Omni / O1)

Within `prompt`, use the `<<<image_N>>>` placeholder to explicitly reference the N-th image in `image_urls` (N starts at **1**). The model substitutes the placeholder with the corresponding image content for:

* Character consistency — `<<<image_1>>>` locks the subject's appearance
* Scene composition — the character from `<<<image_1>>>` appears in the environment of `<<<image_2>>>`
* Multi-asset guidance — embed N reference images at any position

```text Example theme={"system"}
"Take the girl in red from <<<image_1>>> and place her into the snow-mountain scene of <<<image_2>>>, camera slowly zooming in"
```

Corresponding `image_urls`:

```json theme={"system"}
"image_urls": [
  "https://cdn.example.com/girl-red.jpg",
  "https://cdn.example.com/snow-mountain.jpg"
]
```

<Note>
  Non-Omni / O1 models (v2.6 / v3) do not parse `<<<image_N>>>` placeholders and forward them to the model as literal text.
</Note>

## Resource Limits

| Item                           | Limit                                            |
| ------------------------------ | ------------------------------------------------ |
| Reference image (per file)     | ≤ 30MB, JPG / PNG / WEBP                         |
| Reference image (count)        | I2V: 1; Omni / O1 multimodal: ≤ 4 recommended    |
| Reference video (`video_list`) | MP4 / MOV, ≤ 100MB, 2-30 sec, **at most 1 clip** |
| Prompt                         | ≤ 2500 characters                                |
| Output                         | MP4, link valid for 24 hours                     |

## Related Docs

* [Task System Reference](/en/docs/task-system) — task state machine / polling cadence / async push
* [Request & Response Format](/en/docs/request-response) — common error codes / headers / rate limits
* [Authentication](/en/docs/authentication) — API key signup and usage