Aliyun Bailian | 阿里云百炼 icon

Aliyun Bailian | 阿里云百炼

Access Alibaba Cloud Aliyun Bailian video generation, editing, and task APIs.

Video → Reference to Video | 参考生视频(wan2.6)

AI-generated

Summary

Generate a new video clip based on provided reference video URLs and a text prompt using the Wan 2.6 reference-to-video model.

Inputs

  • referenceModel (required) — Select the reference-to-video model (only Wan 2.6 Reference-to-Video available).
  • referenceVideoUrls (required) — Provide one to three HTTP(S) video URLs, each on its own line, that serve as references for generation.
  • referencePrompt (required) — Text prompt to guide the content and style of the generated video.
  • referenceNegativePrompt — Optional text describing elements to avoid in the generated clip (e.g., blurry, low quality).
  • referenceSize — Output video resolution selected from a list or specified by expression, depending on the model.
  • referenceDuration (required) — Requested duration of the generated clip in seconds; options are 5 or 10 seconds.
  • referenceShotType (required) — Select single shot or multi shot mode for the generated video.
  • referenceAudio — Boolean to indicate if audio should be generated along with the video when supported by the model.
  • referenceAdvancedOptions — Advanced settings including embedding AI watermark, setting a deterministic seed, or randomizing the seed for generation.
  • waitForTask — Boolean to determine if the node should poll and wait until the generation task completes or return immediately with the task ID.
  • pollInterval — The interval in milliseconds between polling attempts while waiting for task completion; only applies if waitForTask is true.
  • maxWaitSeconds — The maximum time in seconds to wait while polling for task completion before timing out; only applies if waitForTask is true.

Output shape

a JSON object containing the generated video's metadata and URLs, or the task ID if waiting is disabled

If 'Wait for Task Completion' is true, the node polls the task until it succeeds or fails and returns the final result. Otherwise, it returns immediately with the task ID for manual tracking. Outputs include generated video URLs and metadata aligned with the selected model and options.

Links

Discussion