Seedance icon

Seedance

在 n8n 中创建、查询和管理 Seedance 2.x 视频任务，并使用 Seedream 5.0 lite 生成图片。注意：生成结果 URL 默认仅 24 小时有效，请及时下载转存。

Actions3

创建任务

AI-generated

Summary

Create a Seedance video task from a text prompt, frame images, or multimodal reference materials, or generate images with Seedream 5.0 Lite. The operation submits the request and returns task details for video generation or image metadata with generated images attached as binary data.

Inputs

  • 生成模式 (required) — Choose video generation (default) or image generation.
  • 视频模型与创建模式 (required) — For video, select Seedance 2.5, 2.0, or 2.0 Fast and choose text-to-video, first-frame, first-and-last-frame, or multimodal reference generation. Text-to-video requires a prompt; frame modes require their respective images. Multimodal mode requires at least one reference image, video, or audio. Images and audio can come from public URLs, input binary properties, or the Volcengine Ark asset library; videos can come from URLs or the asset library. Binary frame inputs use the configured property name, defaulting to data.
  • 视频参数 — Set resolution, aspect ratio, duration, audio generation, and (for Seedance 2.5) MP4 or MOV output. Available values depend on the selected model and task; Seedance 2.5 frame modes and multimodal editing or extension use adaptive aspect ratio. Video editing uses the input video's duration; reference generation and extension allow duration selection.
  • 高级选项 — Optionally enable a watermark or return the generated video's last-frame URL in task query results, and set task expiration from 1 to 72 hours.
  • 图片操作与提示词 (required) — For image generation, choose text-to-image or image-to-image and provide the required image prompt. The model is Seedream 5.0 Lite; prompt optimization is enabled by default.
  • 图片参考图 — For image-to-image, provide reference images by public URL, Base64, or input binary property. Multiple values can be comma-separated, or added individually; up to 14 references are supported.
  • 图片输出选项 — Choose 2K or 3K resolution and an aspect ratio. Optionally enable related-image generation (up to 15 images, with references plus requested outputs capped at 15), add a watermark, or enable web search.

Output shape

One output item per input item. Video mode returns a JSON object with taskId, status, optional createdAt, requestSummary, and the raw API response. Image mode returns JSON with requestSummary, per-image success or error metadata, and optional usage and tool-call data; each successful image is attached as binary image1, image2, and so on.

Video creation submits an asynchronous task; it does not wait for completion or download the video. Image binary outputs contain PNG data. If every returned image fails, the operation throws an error. With Continue On Fail enabled, errors are returned as an item containing error and status: "failed".

Links

Discussion