generate-videolisted

# Generate Video ## Catalog - Display name: Generate Video - Category: Video and audio - Status: Ready - Route label: Guide-routed cataloged video model - Tone: purple - Short description: Generate one coherent video from a shot brief using a deliberately selected current model and one observable action. ## What this skill does - Turns text into one shot with explicit subject, action, environment, camera, duration and sound. - Selects a current model from capabilities, not from the phrase text-to-video. ## How to use Provide subject, one action, setting, camera, duration, ratio, style and audio preference. #### Example ```text Generate a 5-second 16:9 video of a paper boat drifting through a rain puddle; slow low tracking shot; natural rain audio. ``` ## Workflow 1. Compile one atomic shot with subject, action, setting, camera, light, duration, ratio and audio. When the request is underspecified, choose one visible change from A to B with a clear emotional or visual payoff; avoid a merely drifting camera over a static subject. 2. When references are supplied, label each role as identity/content, first or last frame, motion, style, or audio timing. Reject a camera move that fights the source composition—for example, revealing space outside a tightly cropped reference—unless an upstream expansion step is approved. 3. Select exactly one cataloged model before submission: MiniMax H3 for large-amplitude/natural human motion or H3 multimodal and first/last-frame modes; Kl