Seedance Video Storyboard Prompt

Problem it solves

When producing short-video scripts, the hard part is often not “having copy,” but keeping copy, shots, dialogue, effects, and timing aligned. seedance-storyboard turns a topic or spoken-script draft into an executable storyboard table: it classifies the video type, selects a matching copywriting model, and outputs shot-level details such as framing, camera movement, visuals, voice-over, captions/effects, and duration, then adds prompts that can go into Jimeng or Seedance.

How it works

The skill routes requests by video type:
- Avatar talking-head: uses the ODCR copy model and follows Path A, optimized for Jimeng digital-human and talking rhythm.
- Corporate, product, and e-commerce: uses VIPS, PASU, or HSPC respectively, follows Path B, and generates Seedance 2.0 shot prompts.

The main steps are:
1. Detect the video type and input, then dispatch to the matching workflow.
2. Load the bundled character and scene cards to complete the host, voice, location, and lighting setup.
3. Output a 7-column storyboard table with shot ID, framing, camera movement/angle, visual description, voice-over, captions/effects, and duration.
4. Estimate duration at roughly 5.5 characters per second, split into Jimeng sub-segments, and rebalance clips to avoid exceeding per-segment limits.
5. Run quality checks for character count, row duration, clip balance, prompt completeness, visual logic, and copy-model coverage.

Boundaries

This skill is best used for pre-production short-video scripting, not direct video generation. The input should include a topic, spoken copy, target style, or registered roles. For very long copy, it should first flag limits and propose compression or segmentation. If role settings are missing, matching falls back to default rules and may feel less specific. Path B depends on Seedance 2.0 constraints, so it works best with context such as seedance2-skill and daily-koubo-generator.