What problem does it solve? Long stories, campaigns, and connected scenes exceed what a single Seedance 2.0 generation can reliably produce, and naive clip-by-clip prompting breaks continuity of characters, locations, and narrative state. This Skill plans the whole story globally while compiling only the next unresolved clip prompt, keeping identity and action consistent across generations. ## Core Features & Use Cases - Sequence Classification: Detects when a request is a multi-clip project versus a standalone clip, and classifies generation mode (T2V, I2V, V2V, R2V, FLF2V), sequence relation, shot structure, and surface profile. - Scene and Clip Architecture: Groups beats into scenes with chain-depth caps, canonical anchor references, arc positions, and audio plans, then divides scenes into generation-sized clips each with one narrative job and felt intent. - Stateful Continuity Management: Tracks planned versus observed start/end states, continuity locks, and a Project State Capsule so each new prompt matches what Seedance actually produced. - Use Case: A user asks for a five-part product launch video. The Skill outputs a story spine, scene map, Clip 01 contract and prompt, provisional intent cards for later clips, and instructions to return each generated clip before the next prompt is finalized. ## Quick Start Ask the agent to plan my connected multi-clip Seedance video story and generate the first clip prompt using the seedance-sequence skill.
