doublesq97-ui/su-ai-short-drama / skills/ai-short-drama-studio-female
ai-short-drama-studio-female
VerifiedDevelop and direct female-audience AI short dramas through role-specific workflows for showrunning, screenwriting, casting and original character design, art, performance, cinematography, storyboarding, sound, post, continuity, and asset handoff. Use when the user asks for 女频短剧、微短剧、言情或非言情女性向故事、女主成长、关系戏、职业戏、女性群像、选题、剧本、改稿、选角定妆、分镜运镜、生图/生视频制作说明、项目圣经或跨工种协作。
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.5 KB |
| agents/openai.yaml | config | 290 B |
| assets/templates/character-card.md | markdown | 1.9 KB |
| assets/templates/development-review.md | markdown | 2.8 KB |
| assets/templates/episode-script.md | markdown | 1.2 KB |
| assets/templates/project-bible.md | markdown | 4.0 KB |
| assets/templates/scene-card.md | markdown | 1.5 KB |
| assets/templates/storyboard-handoff.md | markdown | 1.5 KB |
| references/art-performance-sound.md | markdown | 10.9 KB |
| references/casting-character-design.md | markdown | 10.0 KB |
| references/cinematography-storyboarding.md | markdown | 10.0 KB |
| references/continuity-assets.md | markdown | 9.6 KB |
| references/female-story-engine.md | markdown | 22.8 KB |
| references/roles-and-handoffs.md | markdown | 5.9 KB |
| references/screenwriting-core.md | markdown | 22.8 KB |
| references/showrunning.md | markdown | 6.6 KB |
| references/tool-adapters.md | markdown | 11.3 KB |
references/art-performance-sound.md
View on GitHub# Art direction, performance, sound, and post
Use this reference when a production needs a coherent scene world, playable behavior, and a sound/post plan. Keep the three crafts separate even when one agent performs all of them.
## Shared rule
Work from locked story intent and approved assets. Each craft may expose a defect but must return structural changes to the owning role. Do not use visual polish, exaggerated acting, or loud sound to compensate for unclear causality.
## Art and location direction
### Ownership
The **art / location director** owns spatial design, palette, material, set dressing, practical action support, props, environmental states, and visual continuity. This role does not choose camera coverage or change the dramatic action to fit a decorative image.
### Required inputs
- locked scene action, time, place, period, and emotional function;
- character movement, entrances, exits, contact, props, and safety needs;
- approved character look states and production format;
- recurring uses of the location and planned state changes;
- cultural, professional, legal, medical, military, or historical accuracy needs;
- source and authorization status of visual references.
### Scene-design workflow
1. **State the dramatic function.** Define what the space enables, restricts, reveals, hides, or symbolizes without relying on vague mood words.
2. **Draw the playable space.** Record entrances, exits, vertical levels, furniture, hazards, sightlines, cover, working surfaces, and character routes.
3. **Set permanent anchors.** Choose the few structural elements that must identify the location across views and time states.
4. **Build palette and material logic.** Derive color, texture, wear, maintenance, and light response from period, climate, economics, occupation, and story state.
5. **Register interactive props.** Give every handled, transferred, broken, consumed, opened, read, or weaponized object an ID and before/after state.
6. **Define scene variants.** Record day/night, weather, power loss, ceremony, damage, occupation, cleanup, or seasonal states without changing fixed geography.
7. **Test for action.** Walk the scripted movement through the design. Revise any layout that blocks camera, performance, contact, or continuity.
8. **Create bounded references.** Separate spatial layout, material, lighting, atmosphere, prop, and graphic references so downstream tools know each responsibility.
### Art-direction rules
- A location reference should explain space, not merely provide a beautiful close-up.
- Preserve relative positions of doors, windows, stairs, furniture, fixtures, and signature props.
- Let status and history emerge from condition, access, scale, organization, and use—not from unlicensed luxury logos.
- Use exact signage, documents, interfaces, packaging, and branding only when rights and post-production methods are established.
- Avoid filling every surface. Leave playable negative space for bodies, camera, captions, and action.
- Treat fire, weapons, medical equipment, stunts, vehicles, animals, water, and electrical damage as specialist or high-risk elements.
### Art deliverable
Provide the locked scene card, spatial map, state variants, palette/material board, prop register, visual-reference responsibilities, authorization status, and a list of continuity risks.
## Performance direction
### Ownership
The **performance director** owns intention, obstacle, tactic, subtext, playable action, blocking behavior, reaction, emotional transition, and relationship-specific variation. This role does not prescribe lenses or use personality adjectives as a substitute for action.
### Build a playable beat
For each beat, specify:
```text
Objective: what the character wants now
Obstacle: what prevents it
Tactic: what the character does to the other person or environment
Visible action: behavior the camera can observe
Subtext: what is withheld or contradicted
Trigger: what changes the tactic
Exit state: body, gaze, emotion, relation, and unfinished impulse
```
Prefer verbs such as test, conceal, corner, reassure, provoke, delay, recruit, inspect, deflect, protect, or dominate. Avoid “act emotional,” “be cool,” or “show chemistry” without playable behavior.
### Action-chain discipline
Write complete, motivated sequences. A phone call may require noticing vibration, checking the caller, choosing whether to answer, changing posture, answering, listening, and reacting. A drink may require reaching, grasping, lifting, pausing, sipping, swallowing, and setting down.
Do not add random fidgeting merely to prevent stillness. Secondary action must come from objective, environment, habit, status, or emotional regulation. Stillness is valid when it is active listening, concentration, refusal, threat, shock, concealment, or a deliberate power move.
### Blocking and contact
- Record start position, facing, eyeline, distance, route, contact, and end position.
- Name who initiates touch, what body area or object is involved, the force, duration, and release.
- Preserve personal-space changes as relationship information.
- Give background performers simple purposes and routes; do not leave them frozen, but do not overload them with detailed micro-actions.
- For crowds or fights, prioritize readable group behavior and action geography over simultaneous individual acting.
### Emotional continuity
Track emotion as a change in strategy and body state, not a sequence of labels. Record:
- incoming level of tension, fatigue, trust, injury, intoxication, or concealment;
- the event that changes the character;
- the smallest visible evidence of that change;
- the outgoing state the next shot must inherit.
Protect reaction time after consequential information. Avoid resetting the face and body at every cut.
### Performance and generated media
- Keep one dominant action and one readable emotional transition per difficult generation segment.
- Simplify overlapping contact, hand work, eating, crowd reactions, or fast dialogue when the verified tool cannot sustain them.
- Use approved identity and performance references only. Do not use a real person's unauthorized acting footage or voice as a motion, face, or style shortcut.
- A generated performance remains a candidate until take selection and continuity review.
### Performance deliverable
Provide a beat sheet, blocking notes, contact details, incoming/outgoing states, relationship exceptions, specialist needs, and a short list of behavior that must not drift.
## Sound and post coordination
### Ownership
The **sound and post coordinator** owns the sound-layer plan, voice strategy, dialogue replacement, ambience, effects, music, OS treatment, captions, visual text, editorial dependencies, and final handoff. The role decides where an element should be produced only after verifying the current tool chain.
### Separate the layers
Plan each layer independently:
| Layer | Purpose | Typical source | Continuity record |
|---|---|---|---|
| Production dialogue | Spoken scene action | recorded, generated, or replaced | speaker, take, language, timing, rights |
| OS / narration | Non-diegetic thought or narration | dedicated voice workflow | narrator, perspective, processing, timing |
| Room tone / ambience | Place, scale, weather, activity | captured, generated, library, or designed | scene state and loop length |
| Foley | Body and object contact | recorded, library, or designed | action and sync point |
| Designed effects | Story emphasis, powers, technology, impact | licensed library or design | cue, motif, intensity, version |
| Music | Structure, emotion, motif, release | original or licensed | cue, rights, edit, stem |
| Captions / visual text | Accessibility and exact information | post-production | language, safe area, style, version |
Do not assume a visual generator should also create dialogue, music, or effects. Do not assume it cannot. Verify capabilities and quality, then assign each layer deliberately.
### Dialogue and voice plan
- Preserve meaning, speaker identity, pronunciation, breath, interruption, and relationship-specific rhythm.
- Read dialogue aloud and test intelligibility against action and ambience.
- Record or synthesize separate clean dialogue when production output is inconsistent or unusable.
- Track a voice identity by approved performer or original synthetic profile, language, pronunciation guide, emotional range, and version.
- For a private person, do not clone or imitate a recognizable voice without clear authorization and consent for the intended synthetic use.
- This workflow does not support a recognizable replica of a public figure, celebrity, politician, performer, or creator, even through samples, stylization, feature mixing, or indirect descriptors. Use an original voice profile instead.
### OS, captions, and exact text
OS and narration must be distinguished from spoken dialogue by perspective, timing, and treatment. Captions, messages, documents, UI, logos, and other exact text should be authored and verified in post unless a verified method can produce them reliably. Keep accessibility captions faithful to spoken content; do not use decorative text as a replacement.
### Music and effects
- Give each cue a story function: establish, anticipate, accelerate, interrupt, reveal, release, or connect a motif.
- Leave room for dialogue and important production sound.
- Use silence and room tone intentionally; do not cover every beat with score.
- Track licenses, edit permissions, territory, term, and required attribution for music and sound libraries.
- Design impact through preparation, transient, body or object response, and aftermath rather than only greater loudness.
### Content safety in sound and image
When direct violent, medical, self-harm, or traumatic detail is unnecessary or inappropriate, communicate consequence through framing, obstruction, off-screen sound, reaction, aftermath, and restrained effects. Do not relabel prohibited content, disguise graphic material, or alter prompts to bypass a provider's policy. Follow the provider's safety requirements and the project's rating.
### Sound and post deliverable
Provide:
- dialogue and voice-source plan with authorization status;
- scene ambience and room-tone map;
- Foley, effects, OS, narration, music, caption, and visual-text cue sheets;
- sync points, replacement tasks, stems, file versions, and known gaps;
- editorial notes for rhythm, transitions, reaction holds, and inserts;
- rights records and required attributions.
## Cross-craft integration check
Before handoff, confirm:
- the scene layout supports every blocking and camera route;
- props have consistent owners, positions, states, and sounds;
- performance states survive shot and edit boundaries;
- ambience matches location, time, weather, and crowd state;
- exact text, voice, music, logos, and references have a rights-cleared route;
- unsafe or graphic material is handled through responsible storytelling, not policy evasion;
- each unresolved issue has one owning role and one next action.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |