Create a video generation
Recommended flow: first call POST /v1/videos/prepare to choose and validate the model settings and review the preliminary estimated_price. Then send those same settings plus the final prompt in this request. The prepare endpoint does not accept a prompt or start a generation. This endpoint starts an asynchronous video generation; no defaults are applied to required settings. Validation checks resolution, then aspect_ratio for that resolution, then size for the selected pair; a 400 validation detail includes allowedValues. Use GET /v1/models?type=video to find available models, poll generation_id until completed, then fetch content_url. The MP4 artifact is retained in R2 for 30 days.
Authentication
API key passed as Authorization: Bearer token.
Headers
Request
Video model ID. Find supported IDs using GET /v1/models?type=video.
Optional image or video inputs. Supply owned file IDs; external URLs are not accepted.
