Extend a video

Continue one video from its last frame and get ONE longer clip: the source, then the new footage. Send exactly one source (`video_url` or `source_generation_id`), a `prompt` for what happens next, and `duration_seconds` for the added length (one of the model's own lengths). The default model is seedance-2.5 (the animate-still role). Any image-to-video model is allowed: kling-4, seedance-2.5, gemini-omni-flash, wan-2.7, veo-3.1, seedance-2, seedance-2-pro, grok-imagine, h3-max, flux-3-max. The shape follows the source, so there is no `aspect_ratio`. When you send no `resolution`, the model resolution closest to the source is used. The server makes two ordinary jobs: the continuation (this response) and, when the continuation completes, one merge that joins the source and the continuation. The quote is both: the continuation plus one merge credit, in `credits_breakdown`. The source plus the continuation must fit in 180 seconds. A `webhook_url` is delivered for the stitched clip. Read the run, and the stitched clip, at `/v1/workflows/extend/{id}` with the job id this call returns. If the merge cannot start, the continuation stays charged and usable, and the stage is `merge_failed`.

curl -X POST "https://example.com/v1/workflows/extend" \  -H "Idempotency-Key: stringstringstri" \  -H "Content-Type: application/json" \  -d '{    "prompt": "string"  }'

{  "data": {    "kind": "quote",    "quote_id": "3c071a1d-db86-46a7-9dc8-72ba3fbca992",    "credits": 0,    "expires_at": "string",    "pricing_version": "string",    "confirmation": {      "property1": null,      "property2": null    }  }}