tfcbot Director — Clip Gen
Vidjutsu supports one clone-video model: Kling 3.0 Motion Control.
Submit
POST https://api.vidjutsu.ai/v1/clones/video
Authorization: Bearer <VIDJUTSU_API_KEY>
Idempotency-Key: <stable workflow key>
Content-Type: application/json
{
"startingImageUrl": "<generated starting image URL>",
"sourceVideoUrl": "<staged source video URL>",
"model": "kling",
"prompt": "<optional concise motion-control guidance>"
}
The source must be public HTTPS and 3–15 seconds. Acceptance is asynchronous:
{"id":"<Vidjutsu clone task ID>","status":"processing"}
Persist id immediately. Reusing the same idempotency key prevents duplicate
generation when the original request outcome is replayed.
Read status
GET https://api.vidjutsu.ai/v1/clones/video/{id}
Authorization: Bearer <VIDJUTSU_API_KEY>
processing: checkpoint the task ID and check again later.completed: use the returnedvideoUrl.failed: stop and report the returnederror.
Status reads use Vidjutsu-owned state updated by authenticated runner callbacks. They do not poll Kling and do not expose an upstream task ID.
Rules
- Never select or fall back to another model or Kling version.
- Never call the dedicated runner or an upstream provider directly.
- Never resubmit merely because a task is still
processing. - Preserve the returned video without resizing or transcoding until post-production.
- Clone submissions share the 50/day clone admission group; status reads are unmetered and have no separate per-operation charge.