What problem does it solve? Producing cinematic AI video clips requires mastering prompt structure, camera language, multi-shot continuity, and provider configuration. This Skill encodes proven Seedance 2.0 prompting methodology and routes all generation through JiubanAI's managed video service, so you never handle Volcengine credentials or polling logic yourself. ## Core Features & Use Cases - Structured prompt engineering: Apply the Higgsfield methodology with shot-structure openers, temporal beat markers, VFX brackets, and camera negation rules for photorealistic multi-shot output. - Multi-clip continuity: Chain clips using first_clip references so later segments continue from the previous clip's final state instead of replaying the opening action. - Identity-locked reference video: Use reference-to-video with [identity_lock] blocks and anti-drift phrasing to keep characters consistent across shots. - Use Case: A content team needs a 10-second vertical product teaser with synced audio and consistent character appearance. The Skill generates the clip via video_selector at 9:16, 720p, with an identity-locked prompt, then verifies motion, lip-sync, and grade against its checklist. ## Quick Start Ask the agent to generate a 10-second 16:9 cinematic video clip from your prompt using the video_selector tool with the JiubanAI Seedance runtime.
