Generate a video using an AI video generation model. Supports text-to-video and image-to-video (when a single reference image is provided). Only one reference image is supported — provide exactly one image path or omit images entirely for text-to-video. Generation can take up to 1 hour to complete (provider queue stalls are common — do not abandon a job that is still pending). Returns the path to the generated video file (the result also includes a "Video content:" description of what the clip shows) — use [VIDEO:path] in your reply to send it to the user.
Reference image limits: the reference must be a PNG, JPEG, or WebP file. References larger than 1.5 MB are automatically compressed (downscaled and re-encoded as JPEG) to fit, and a reference near the cap may be compressed further so the total request stays within the provider's body limit. Missing, unreadable, or undecodable references are hard errors — the tool fails visibly rather than falling back to text-to-video, so fix or remove the reference and retry.