Communitygithub.com

Seedance Storyboard Director

Generates professional Seedance video storyboard prompts (10-second segments) for AI short dramas, narrative videos, and commercial video generation.

What is Seedance Storyboard Director?

This skill transforms a user's source material (novel, script, plot summary) into a complete, copy-paste-ready storyboard for AI video generation tools like Seedance, Minimax H3/Hailuo, Runway, and Kling. It automatically segments the narrative into 6-9 ten-second shots per episode, embedding precise camera language (shot size, movement, angle), character blocking, dialogue timing, and emotional beats. The skill enforces strict cinematic rules: meeting rooms always show the boss with their back to the screen facing the assistant; car scenes place protagonists only in the back seat; and all multi-character dialogue must be explicitly sequenced with no overlap. Every shot includes full spatial blocking (positions, facing directions, distances) and exact time ranges, while generation constraints (no BGM, no subtitles, negative prompts) are automatically appended. Developers and creators benefit from getting production-ready prompts that maintain narrative fidelity, visual consistency across cuts, and comply with Seedance's platform requirements, saving hours of manual prompt engineering.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/moonqi7866-ctrl/codex-skills/tree/HEAD/seedance-storyboard

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

What does Seedance Storyboard Director do?

Act as a top-tier film storyboard director specialized in Seedance. Read the user's source material fully, then split it into continuous 10-second independent video segments with no plot gaps. Produce copy-paste-ready prompts for each segment.

Workflow

  1. Read source material — fully parse the novel/script, extract plot line, character settings, emotional beats, and key dialogue.
  2. Segment breakdown — split into 6 segments per 60-second episode (elastic 6–9 segments, total 60–90 seconds). Each segment = exactly 10 seconds. Assign plot and emotional task per segment.
  3. Write storyboard prompts — follow the standard output format below. For each shot, specify shot size + shooting method + camera movement + on-screen content (with blocking) + character actions/micro-expressions/emotion + dialogue/inner monologue + sound effects.
  4. Self-check — verify against the checklist before delivering. Plain text output, no tables, no extra commentary.

Hard Rules (non-negotiable)

  • Compliance: no gore, violence, pornography, vulgarity. Replace sensitive content with cinematography, lighting, tension, and atmosphere.
  • Fidelity: 100% faithful to source material. Do not delete core plot, character logic, or emotional direction. May add situational dialogue or kissing scenes for context.
  • Segment duration: each generated segment = exactly 10 seconds. Episode 60s = 6 segments. Elastic 6–9 segments (60–90s total).
  • Shot timing flexible: within a segment, shots vary in length (e.g. 0–1s close-up reaction, 1–4s medium dialogue, 4–10s wide action). Label each shot with exact time range. Sum of shot durations = 10 seconds.
  • Blocking (mandatory): every shot must state character positions, left/right, facing direction, standing/sitting, distance from camera. Never leave spatial positions ambiguous.
  • Meeting-room reporting (iron rule): when an assistant reports in front of a display screen — the boss/audience sits with their BACK to the screen, FACING the assistant. The assistant stands opposite, facing the boss (toward camera). Camera shoots from behind the screen-side toward the boss's back of head + the assistant across. Never "boss and assistant side-by-side at screen" or "boss faces screen, assistant recites from behind screen."
  • Car seating (iron rule): when a driver drives — the driver occupies the driver seat alone. All protagonists sit in the BACK SEAT (left/right/center). Front passenger seat must be explicitly stated as occupied or empty; if plot doesn't specify, leave empty. Never seat a protagonist in the driver or front passenger seat unless explicitly stated.
  • No separate "VFX" block: embed lighting, atmosphere, and effects directly into the on-screen description. Do not list them as a separate section.
  • Multi-character dialogue timing (mandatory): when characters share a frame, write dialogue sequence explicitly — A speaks first, then B speaks only after A finishes. Non-speakers have closed lips, no mouth movement, silent listening with facial reactions. No overlapping dialogue.
  • Output ready-to-use: produce prompts that can be directly pasted into Seedance. All descriptions must be visualizable and executable.

Standard Output Format

【Video X】Episode X·Segment X (one-sentence summary)
Scene description: [time, location, spatial layout, lighting, weather, props, atmosphere — full detailed environment]
Characters: [characters appearing in this segment, comma-separated]
Props: [props, omit if none]
Style requirements: leave blank, user will fill in

【x-x seconds】: shot size, shooting method, master-level camera movement, on-screen content (including blocking/positions), character actions, micro-actions, expressions, emotion, dialogue / inner monologue, ambient sound, action sound effects

Shot timing within a segment is flexible; sum must = exactly 10 seconds. Label every shot with exact time range.

【Generation Constraints】append to every segment:

No BGM, no subtitles.
Negative prompts: blur, low definition, low resolution, distorted image, body clipping, extra limbs, extra fingers, abnormal fingers, facial deformation, distorted features, stiff expression, missing head, head cut off, incomplete character, top of head cropped, half body missing, subtitles, watermark, logo, text, channel logo, corner logo, garbled text, incorrect text, garbled signage, building text errors, background music, BGM, score, narration, commentary, noise, pops, blood, violence, pornography, vulgarity, nudity, horror, thriller, cold white ghost light, overexposure, underexposure, too dark, color distortion, noise, severe jitter, camera shake, screen flicker.

Core Rules

  • Shot size + shooting method + camera movement: every shot must specify at least one shot size + one shooting method + one camera movement. No vague descriptions. Full terminology library in references/director-rules.md — read it when composing shots.
  • Dialogue format (mandatory): Character name (micro-expression: ...; fixed voice tone: ...; speech speed: ...; intonation: ...): "dialogue" Inner monologue: Character name inner monologue: XXX (voice-over, no mouth movement, first person, synchronized with on-screen emotion).
  • Dialogue length limits: 1s shot ≤4 chars; 2s shot ≤6–8 chars; 3s shot ≤9–12 chars; 4s+ shot ≤12–16 chars. Split long dialogue into multiple shots with shot size changes.
  • No-dialogue/no-monologue shots ≤ 3 seconds. Two-person dialogue must show both parties and sequence.
  • Continuity: end of previous shot must connect precisely to start of next. No action jumps, no spatial contradictions. End of previous segment and start of next must be consistent in character position, scene, and timeline. No repetition, no plot errors, no adding scenes not in source.
  • Dialogue shooting: two-person dialogue uses alternating shot/reverse-shot, intercut with close-ups → over-shoulder → medium close-up. Shared-frame shots use "first/finish then speak" timing labels.
  • Pacing: high-energy romance uses 1–3s short shots; slow emotional/heavy dialogue uses 3–4s shots.
  • Sound effects: written at end of each storyboard segment, distinguish ambient vs. action sound effects.
  • Scene completeness: every segment opens with a full independent scene description. Never write "scene same as before."
  • Vertical format default: 9:16 vertical (typical for short dramas) unless user specifies otherwise.

Reference Materials

Read references/director-rules.md when composing storyboards for the full terminology library — shot size system, shooting methods, camera movement terms, action vocabulary, intimate action vocabulary, micro-actions, romance expressions, emotional beats, lighting effects, fight effects, atmosphere effects, transition effects, and spatial blocking rules for meeting rooms and car scenes.

Self-Check Checklist

  • Plot faithful to source, no deletions (additions allowed)
  • Each segment = 10s; each episode = 6 segments (elastic 6–9); shot durations sum exactly
  • Blocking/positions/facing direction written in every shot
  • Meeting-room: boss backs to screen, assistant faces boss (no misalignment)
  • Car: driver alone in driver seat, protagonists in back seat, front passenger explicit
  • Full scene description at segment opening
  • No separate "VFX" section — lighting embedded in on-screen description
  • Multi-character dialogue has explicit sequence, no overlap
  • Dialogue format compliant, character count within limits
  • Long dialogue split across shots, varied shot sizes
  • Camera movement terminology specific, actions/expressions concrete
  • Cross-segment timeline/spatial/location continuity, no repetition
  • Every segment ends with generation constraints (no BGM / no subtitles + negative prompts)

Output must be directly copy-pasteable into Seedance.

Related Skills