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

# Wan 2.6 Reference-to-Video

AI-powered reference-to-video service. Generate high-quality video content from reference images or reference videos combined with text prompts. Provides professional video generation capabilities with ready-to-use REST inference APIs, high performance, no cold starts, and affordable pricing.

<Tip>
  This is an **asynchronous** API; only the **task\_id** will be returned. You should use the **task\_id** to request the [**Task Result API**](/docs/de/api-reference/model-apis-task-result) to retrieve the video generation results.
</Tip>

## Request Headers

<ParamField header="Content-Type" type="string" required={true}>
  Supports: `application/json`
</ParamField>

<ParamField header="Authorization" type="string" required={true}>
  Bearer authentication format, for example: Bearer \{\{API Key}}.
</ParamField>

## Request Body

<ParamField body="input" type="object" required={true}>
  <Expandable title="properties" defaultOpen={true}>
    <ParamField body="prompt" type="string" required={true}>
      Text prompt describing the expected elements and visual characteristics in the generated video. Chinese and English are supported. Each Chinese character, letter, or punctuation mark counts as one character, and content beyond the limit is automatically truncated. Use identifiers such as character1 and character2 to reference characters. Each reference video or image should contain only a single character.

      Length limit: 0 - 1500
    </ParamField>

    <ParamField body="reference_urls" type="string[]" required={true}>
      Array of reference image or reference video URLs, used to extract character appearance and voice tone if available. Image count: 0 to 5. Video count: 0 to 3. Total image plus video count must not exceed 5. When multiple reference files are provided, character order follows the array order: the first URL corresponds to character1, the second to character2, and so on. Reference video requirements: duration 1s to 30s, file size up to 100 MB. Reference image requirements: width and height must both be within \[240, 8000] pixels, file size up to 20 MB.

      Array length: 1 - 5
    </ParamField>

    <ParamField body="negative_prompt" type="string">
      Negative prompt describing content that should not appear in the video. It can be used to constrain the video image. Chinese and English are supported. The length must not exceed 500 characters, and content beyond the limit is automatically truncated.

      Length limit: 0 - 500
    </ParamField>
  </Expandable>
</ParamField>

<ParamField body="parameters" type="object">
  <Expandable title="properties" defaultOpen={true}>
    <ParamField body="seed" type="integer">
      Random seed. Valid range: \[0, 2147483647]. If not specified, the system automatically generates a random seed. To improve reproducibility, set a fixed seed value. Note that because model generation is probabilistic, using the same seed does not guarantee identical results every time.

      Value range: \[0, 2147483647]
    </ParamField>

    <ParamField body="size" type="string" default="1920*1080">
      Specifies the generated video resolution in width*height format. Supports 720P options (1280*720/720*1280/960*960/1088*832/832*1088) and 1080P options (1920*1080/1080*1920/1440*1440/1632*1248/1248\*1632).

      Optional values: `1280*720`, `720*1280`, `960*960`, `1088*832`, `832*1088`, `1920*1080`, `1080*1920`, `1440*1440`, `1632*1248`, `1248*1632`
    </ParamField>

    <ParamField body="duration" type="integer" default={5}>
      Generated video duration in seconds. Available values are 5 and 10. The default value is 5. duration directly affects cost: cost = unit price (based on resolution) x duration (seconds). Confirm model pricing before calling the API.

      Optional values: `5`, `10`
    </ParamField>

    <ParamField body="shot_type" type="string" default="single">
      Video generation mode. single: default value, outputs a single-shot video. multi: outputs a multi-shot video. Parameter priority: shot\_type > prompt.

      Optional values: `single`, `multi`
    </ParamField>

    <ParamField body="watermark" type="boolean" default={false}>
      Whether to add a watermark. The watermark is located in the lower-right corner of the video and the fixed text is "AI generated". false: default value, no watermark. true: add watermark.
    </ParamField>

    <ParamField body="prompt_extend" type="boolean" default={true}>
      Whether to enable intelligent prompt rewriting. When enabled, a large language model rewrites the input prompt. This noticeably improves generation quality for shorter prompts, but increases processing time. true: default value, enable intelligent rewriting. false: disable intelligent rewriting.
    </ParamField>
  </Expandable>
</ParamField>

## Response

<ResponseField name="task_id" type="string" required={true}>
  Use the task\_id to request the [Task Result API](/docs/de/api-reference/model-apis-task-result) to retrieve the generated outputs.
</ResponseField>