Overview

This is the video file editing capability Vidu S provides: for a pre-recorded source video, submit a set of scene switches along a timeline, and asynchronously generate an edited MP4. Unlike the real-time editing live stream, this is not streaming — no long-lived connection is established; each submission corresponds to one complete editing task, and the result is obtained by polling or callback.

The four capabilities	Description
Style rendering `style_transfer`	Redraws the video picture in the style of the reference image. Commonly used for anime-style or artistic-style edited footage.
Character replacement `subject_replacement`	Replaces the person in the video with the character from the reference image, preserving the original motion and camera pacing.
Background replacement `background_replacement`	Replaces the video background with the scene from the reference image, while keeping the subject-foreground relationship unchanged.
Virtual try-on `virtual_tryon`	Dresses the person in the video according to the reference image, while preserving their posture and body structure.

Minimal Integration Flow

Create a Task

Submit the source video address and a time-ordered set of scene switches to obtain a task_id for subsequent queries.

curl -X POST "$BASE_URL/s_editing/offline" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "vidu-s2-editing",
    "video": "https://example.com/input/source.mp4",
    "actions": [
      {"time_s": 0,   "image_uri": "https://example.com/input/style.jpg",  "editing_type": "style_transfer"},
      {"time_s": 1.5, "image_uri": "https://example.com/input/clothes.png","editing_type": "virtual_tryon"}
    ]
  }'

The output resolution is 540p, 24fps; width/height and orientation match the source video; the output duration equals the source video duration (it is neither padded nor truncated).

Query the Task and Retrieve the Output

A successful creation only means the task has been accepted, not that the output video has been generated; poll using task_id, or configure callback_url to wait for a callback.

Task status: GET /v2/tasks (can be filtered by task ID); terminal states are success / failed.
Output video: GET /v2/tasks/{task_id}/creations, returns the output video address (plus cover/last-frame images).

Interface Protocol Overview

End-to-End Integration Flow

Call the “Create Task” endpoint, submitting the source video and a time-ordered sequence of scene switches, to obtain a task_id.
Poll the “Query Task” endpoint until it reaches a terminal state of success or failed (you can also configure callback_url to receive a callback instead of polling).
Once success, call the “Query Output” endpoint to retrieve the generated video address.

API Overview

Endpoint	Method and Path	Purpose
POST	`/s_editing/offline`	Create an offline editing task (submit the source video and scene-switch sequence; returns `task_id`)
GET	`/v2/tasks`	Query task status (can be filtered by task ID)
GET	`/v2/tasks/{task_id}/creations`	Retrieve the output video address

General Conventions

Item	Convention
Integration Environment and Authentication	The base URL and authentication follow the existing Vidu MaaS open API (same as other `/s_` and `/v2/` endpoints).
Data Format	`Content-Type: application/json`
Asynchronous Processing and Polling	Tasks are processed asynchronously. A successful creation only means the task has been accepted, not that the output video has been generated; poll the query endpoint, or configure `callback_url` to wait for a callback.
Billing	The create endpoint returns `credits` as `0`; the actual deduction occurs within the task workflow based on source video duration × 1.5, and is reflected in the “Query Task” result.