Boreal

Direction for the visuals, spoken dialogue, sounds, and on-screen text. Structure it with `[VISUAL]`, `[SPEECH]`, `[SOUNDS]`, and `[TEXT]` sections for best results. English is the validated language.

Content or artifacts to avoid in the generated video.

Optional reference image to animate. Guides identity, composition, and appearance; the output follows this image's aspect ratio.

Optional driving audio (narration, dialogue, or a soundtrack). When provided it is preserved in the output instead of generated speech.

Frame ratio. `auto` follows the reference image's own ratio, and is landscape (16:9) when there is no image. A named ratio renders that ratio; with an image it centre-crops the image to it.

Length of the generated video in seconds at 24 FPS, rounded by the model to the nearest legal frame count (5 s -> 121 frames, 10 s -> 241, 15 s -> 361). With `audio_url`, the first `duration` seconds of the audio are used; shorter audio is padded with silence.