Skip to main content
character_orientation is required and determines the output length.
  • image — output is fixed at 5s regardless of duration.
  • video — output matches the reference video’s length (3–30s). The RouterBase playground probes the reference clip and prefills duration; direct API callers should send duration matching what they want billed.
Billing scales linearly with output length at **0.160/s(RouterBase)/0.160/s** (RouterBase) / 0.168/s (list).

Endpoint

Async — returns {id, status: "pending"}; poll GET /v1/videos/generations/{id} until success or failed.

Parameters

"image" | "video"
default:"\"image\""
required
Drives mode selection.
  • image: 5s fixed-length output.
  • video: output length tracks the reference video (3–30s).
string[]
required
Exactly one subject-image URL (HTTPS, PNG / JPEG) — the character to animate.
string[]
required
Exactly one reference-video URL (HTTPS, MP4) supplying the motion.
string
Optional text prompt to bias the generation. Max 2500 characters.
integer
default:"5"
Length of the generated clip in seconds. Range [3, 30]. In image mode this is ignored (output is fixed 5s). In video mode this should match the reference video’s length — used for billing.
boolean
default:"true"
Whether to keep the reference video’s audio track on the output.
string
Content to steer the model away from. Max 2500 characters.

Example

Poll until terminal:

Reference