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

# Seedance 2.0 Video Generation — Create Task API Reference

> POST /v1/kyyvideo2/videos — Create a Seedance 2.0 video task. Pro models need no reference video; standard models require one. Supports multi-modal inputs.

Use this endpoint to submit a video generation task with Seedance 2.0, ByteDance's latest generation video model. Seedance 2.0 supports a rich set of multi-modal reference inputs — you can guide generation using reference images, reference videos, reference audio clips, or by pinning the first and last frames. Two model tiers are available (standard and Pro), each with a fast variant. The API responds immediately with a task `id` — you then poll the [Query Video Task](/api-reference/video/seedance2-query) endpoint until the video is ready.

<Note>
  Video generation is **asynchronous**. The create endpoint returns a task `id` with `status: "queued"`. Use the [Query Video Task](/api-reference/video/seedance2-query) endpoint to check progress and retrieve the final `video_url`.
</Note>

## Base URL & Endpoint

```
Base URL:  https://zcbservice.aizfw.cn/kyyReactApiServer
Endpoint:  POST /v1/kyyvideo2/videos
Full URL:  https://zcbservice.aizfw.cn/kyyReactApiServer/v1/kyyvideo2/videos
```

All requests must include your API key as a Bearer token:

```
Authorization: Bearer YOUR_API_KEY
```

## Supported Models

| Model                   | Reference Video Required | Speed    | Quality  |
| ----------------------- | ------------------------ | -------- | -------- |
| `seedance_2_0`          | ✅ Required               | Standard | Standard |
| `seedance_2_0_fast`     | ✅ Required               | Fast     | Standard |
| `seedance_2_0_pro`      | ⭕ Optional               | Standard | High     |
| `seedance_2_0_fast_pro` | ⭕ Optional               | Fast     | High     |

<Note>
  `seedance_2_0` and `seedance_2_0_fast` **require** at least one reference video (`referenceVideos`). `seedance_2_0_pro` and `seedance_2_0_fast_pro` work without reference videos, making them suitable for pure text-to-video and first/last frame workflows.
</Note>

## Request Parameters

<ParamField body="model" type="string" required>
  The Seedance 2.0 model to use. Accepted values: `seedance_2_0`, `seedance_2_0_fast`, `seedance_2_0_pro`, `seedance_2_0_fast_pro`.
</ParamField>

<ParamField body="prompt" type="string" required>
  A text description of the video you want to generate. Be specific about scene, style, motion, and lighting.

  Example: `"A cat dancing in the rain, cinematic style"`
</ParamField>

<ParamField body="duration" type="integer">
  Output video duration in seconds. Range: `4`–`15`. Default: `5`.

  Longer durations increase generation time proportionally.
</ParamField>

<ParamField body="aspect_ratio" type="string">
  Output video aspect ratio. Default: `16:9`.

  Accepted values: `16:9`, `4:3`, `1:1`, `3:4`, `9:16`, `21:9`, `adaptive`.
</ParamField>

<ParamField body="generateAudio" type="boolean">
  Whether to include synthesised audio in the output video. Default: `true`.

  * `true` — output video includes synchronised audio
  * `false` — output video is silent
</ParamField>

### First / Last Frame Mode

Pin the exact start and/or end frames of the generated video. **Mutually exclusive with `referenceImages`, `referenceVideos`, and `referenceAudios`.**

<ParamField body="first_image" type="string">
  The first frame of the video. Accepts a public URL or an asset library reference in the format `asset://ASSET_ID`.

  * Used alone: image-to-video mode (video starts from this frame)
  * Used with `last_image`: first/last frame mode

  **Image requirements:**

  * Formats: JPEG, PNG, WebP, BMP, TIFF, GIF
  * Aspect ratio (width ÷ height): 0.4–2.5
  * Dimensions: 300–6000 px per side

  Cannot be combined with `referenceImages`, `referenceVideos`, or `referenceAudios`.
</ParamField>

<ParamField body="last_image" type="string">
  The last frame of the video. Must be used together with `first_image`. Same format and dimension requirements as `first_image`.

  Cannot be combined with `referenceImages`, `referenceVideos`, or `referenceAudios`.
</ParamField>

### Multi-Modal Reference Mode

Provide reference media to guide generation style and content. **Mutually exclusive with `first_image` and `last_image`.**

<ParamField body="referenceImages" type="array">
  A list of reference image URLs or asset library references (`asset://ASSET_ID`) to guide visual style.

  * **Count:** 1–9 images
  * **Formats:** JPEG, PNG, WebP, BMP, TIFF, GIF
  * **Aspect ratio:** 0.4–2.5
  * **Dimensions:** 300–6000 px per side

  Cannot be combined with `first_image` or `last_image`.
</ParamField>

<ParamField body="referenceVideos" type="array">
  A list of reference video URLs or asset library references (`asset://ASSET_ID`) to guide motion and content.

  * **Count:** max 3 videos
  * **Resolutions:** 480p or 720p
  * **Duration per clip:** 2–15 seconds
  * **Total combined duration:** ≤ 15 seconds

  The system automatically selects the optimal backend model based on your reference videos.

  Cannot be combined with `first_image` or `last_image`.
</ParamField>

<ParamField body="referenceAudios" type="array">
  A list of reference audio URLs or asset library references (`asset://ASSET_ID`) to guide the generated audio track.

  * **Count:** max 3 clips
  * **Formats:** WAV, MP3
  * **Duration per clip:** 2–15 seconds
  * **Total combined duration:** ≤ 15 seconds
  * **File size per clip:** ≤ 15 MB

  Cannot be combined with `first_image` or `last_image`.
</ParamField>

## Input Mode Mutual Exclusivity

<Warning>
  The two input guidance modes are **mutually exclusive**. You cannot use both in the same request:

  * **First/last frame mode:** `first_image`, `last_image`
  * **Multi-modal reference mode:** `referenceImages`, `referenceVideos`, `referenceAudios`
</Warning>

## Response Fields

<ResponseField name="id" type="string">
  Unique identifier for the video generation task. Save this value — you will use it to poll the [Query Video Task](/api-reference/video/seedance2-query) endpoint.
</ResponseField>

<ResponseField name="object" type="string">
  Object type. Always `"video"`.
</ResponseField>

<ResponseField name="created" type="integer">
  Unix timestamp (seconds) of when the task was created.
</ResponseField>

<ResponseField name="model" type="string">
  The model name you specified in the request.
</ResponseField>

<ResponseField name="status" type="string">
  Initial task status. On successful creation this is always `"queued"`.
</ResponseField>

<ResponseField name="error" type="string">
  Error message. Only present when `status` is `"failed"`.
</ResponseField>

## Code Examples

<CodeGroup>
  ```bash cURL (Text-to-Video) theme={null}
  curl --request POST \
    --url https://zcbservice.aizfw.cn/kyyReactApiServer/v1/kyyvideo2/videos \
    --header 'Authorization: Bearer YOUR_API_KEY' \
    --header 'Content-Type: application/json' \
    --data '{
      "model": "seedance_2_0_pro",
      "prompt": "A cat dancing in the rain, cinematic style",
      "duration": 8,
      "aspect_ratio": "16:9",
      "generateAudio": true
    }'
  ```

  ```bash cURL (Multi-Modal Reference) theme={null}
  curl --request POST \
    --url https://zcbservice.aizfw.cn/kyyReactApiServer/v1/kyyvideo2/videos \
    --header 'Authorization: Bearer YOUR_API_KEY' \
    --header 'Content-Type: application/json' \
    --data '{
      "model": "seedance_2_0",
      "prompt": "Continue the motion from the reference video, cinematic style",
      "duration": 8,
      "aspect_ratio": "16:9",
      "generateAudio": true,
      "referenceImages": [
        "https://example.com/style-reference.jpg"
      ],
      "referenceVideos": [
        "https://example.com/motion-reference.mp4"
      ]
    }'
  ```

  ```python Python theme={null}
  import requests

  url = "https://zcbservice.aizfw.cn/kyyReactApiServer/v1/kyyvideo2/videos"

  headers = {
      "Authorization": "Bearer YOUR_API_KEY",
      "Content-Type": "application/json",
  }

  # Text-to-video with Pro model (no reference video required)
  payload = {
      "model": "seedance_2_0_pro",
      "prompt": "A cat dancing in the rain, cinematic style",
      "duration": 8,
      "aspect_ratio": "16:9",
      "generateAudio": True,
  }

  response = requests.post(url, headers=headers, json=payload)
  data = response.json()

  task_id = data["id"]
  print(f"Task created: {task_id} — status: {data['status']}")
  ```
</CodeGroup>

**Example success response:**

```json theme={null}
{
  "id": "video_fd35ee52-2a98-44a6-b930-29a88ce9b8fd",
  "object": "video",
  "created": 1774836724,
  "model": "seedance_2_0_pro",
  "status": "queued",
  "error": null
}
```

## Tips

<Tip>
  Use `seedance_2_0_pro` or `seedance_2_0_fast_pro` for text-to-video and first/last-frame tasks — these models work without a mandatory reference video, giving you the most flexibility.
</Tip>

<Warning>
  If you are using reference images that contain **human faces or virtual avatars**, you must upload them to the asset library first for content review. Use `asset://ASSET_ID` references rather than direct URLs in those cases.
</Warning>

<Tip>
  Keep your reference videos short (2–5 seconds each) and under the 15-second total cap. The system selects the optimal backend model automatically based on your reference video content.
</Tip>