@milenyumai/film-kit 1.3.0 → 1.4.1

package/README.md

@@ -18,6 +18,7 @@ Film-Kit ships as a single repository with four npm packages:
   - native `.claude/agents/*`
   - cleanup of stale mode-specific Claude artifacts
 - Shared `spatial-blocking` skill for gaze, plane depth, light cohesion, compositing realism, and anti-miniature control.
+- Voice-design aware audio contract with project-level `voiceCast`, shot-level `Audio Plan`, and backward-compatible `Audio direction` blocks.
 - Aligned quality gates across Claude Code, Cursor, Copilot, and Antigravity.
 - Stronger Kling 3.0 and Kling multi-shot guidance, including practical route rules and hard caps.
 
@@ -73,7 +74,7 @@ Best for:
 Key contracts:
 
 - `team-plan.json`
-- specialist validators
+- 8 specialist validators in 3-phase execution (continuity -> parallel quality checks -> delivery)
 - parallel batch generation
 - spatial contract for multi-subject staging
 

package/build/lib/templates.js

@@ -44,6 +44,7 @@ All rules, skills, and workflows are located under \`.agent/\`.
 ### Entry Points
 - **Master Rules:** \`.agent/MASTER.md\` — Complete production ruleset (690 lines)
 - **Architecture:** \`.agent/ARCHITECTURE.md\` — System map & quick reference
+- **Voice Design:** \`.agent/VOICE-DESIGN.md\` — Project-level \`voiceCast\` + shot-level \`audioPlan\`
 - **Model Profile:** \`.agent/model-profile.md\` — Active model rules and constraints
 - **Agent:** \`.agent/agents/prompt-engineer.md\` — Senior prompt engineer agent
 
@@ -71,7 +72,7 @@ All rules, skills, and workflows are located under \`.agent/\`.
 ## Goal
 When the user asks \`/generate\`, convert the scenario into:
 - \`${config.outputDir}/project-info.md\` — Characters, settings, arc mapping
-- \`${config.outputDir}/shot-plan.json\` — Single-agent plan + policy contract
+- \`${config.outputDir}/shot-plan.json\` — Single-agent plan + policy + \`voiceCast\` contract
 - \`${config.outputDir}/shots/SHOT01.md, SHOT02.md, ...\` — Production shot files (with coverage included)
 - \`${config.outputDir}/reports/SAFETY-REPORT.md\` — Safety gate result
 - \`${config.outputDir}/reports/DELIVERY-REPORT.md\` — Delivery gate result
@@ -88,6 +89,7 @@ Each \`SHOTNN.md\` is a **single file** containing ALL shot details:
 - 🔗 Chain status (FIRST / CHAINED / CHAIN BREAK)
 - İLK FRAME (start frame image prompt — min 60 words)
 - SON FRAME (end frame image prompt — min 60 words)
+- AUDIO PLAN (machine-readable shot audio contract)
 - VİDEO (video prompt with audio direction — min 80 words)
 - COVERAGE SHOTS (2-3 coverage shots within same file — each min 60 words)
 - 🇹🇷 Turkish summary for each section
@@ -99,6 +101,7 @@ Each \`SHOTNN.md\` is a **single file** containing ALL shot details:
 - **AUTO-SAFETY:** Proactively reframe content that may trigger safety filters
 - **Frame Chaining:** Last frame of SHOT[N] = First frame of SHOT[N+1]
 - **Coverage Mandatory:** Every main shot includes 2-3 coverage sub-shots in same file
+- **Voice Design:** \`shot-plan.json\` keeps top-level \`voiceCast\`; every speaking VIDEO section keeps \`Audio Plan\`
 - **Music: NONE** by default (user must explicitly request)
 - **SLOW BURN:** 8-second duration, split actions into multiple shots
 `;
@@ -111,6 +114,7 @@ function buildCursorLegacyRules(config) {
 ## CRITICAL: SKILL LOADING PROTOCOL
 Before generating ANY prompts, read the appropriate skill files from .agent/skills/:
 Read .agent/model-profile.md first to apply model-specific rules.
+Read .agent/VOICE-DESIGN.md when dialogue, narrator VO, or reusable speaker identity exists.
 
 | Skill | Path | When |
 |-------|------|------|
@@ -135,6 +139,7 @@ Read .agent/model-profile.md first to apply model-specific rules.
 9. Frame chaining: Last frame of SHOT[N] = First frame of SHOT[N+1].
 10. ILK/İLK FRAME section must contain a code block even for chained shots.
 11. ONE FILE PER SHOT: Each SHOTNN.md contains main shot + all coverage shots.
+12. Keep top-level \`voiceCast\` in ${config.outputDir}/shot-plan.json and \`Audio Plan\` in every speaking VIDEO section.
 
 ## WORKFLOWS
 - /generate → Read .agent/workflows/generate.md
@@ -164,6 +169,7 @@ alwaysApply: true
 ## Entry Point
 Read \`.agent/MASTER.md\` for complete production ruleset.
 Read \`.agent/ARCHITECTURE.md\` for system overview.
+Read \`.agent/VOICE-DESIGN.md\` for voice identity and shot audio contracts.
 Read \`.agent/model-profile.md\` for active model constraints.
 
 ## SKILL LOADING (MANDATORY)
@@ -187,6 +193,8 @@ All skills at: \`.agent/skills/[name]/SKILL.md\`
 ## CRITICAL RULES
 - AUTO-ANONYMOUS: Replace ALL real names with physical descriptions
 - Dialogue naming follows \`${config.outputDir}/shot-plan.json\` policy
+- \`shot-plan.json\` stores top-level \`voiceCast\`
+- Every speaking VIDEO section includes \`Audio Plan\`
 - AUTO-SAFETY: Proactively reframe sensitive content
 - Frame chaining: Last frame SHOT[N] = First frame SHOT[N+1]
 - Coverage: 2-3 sub-shots per main shot (min 60 words each, in same file)
@@ -216,8 +224,9 @@ Single-agent shot package generation. Claude Code may use the native \`prompt-en
 ## Mandatory Read Order
 1. \`.agent/model-profile.md\`
 2. \`.agent/MASTER.md\`
-3. \`.agent/ARCHITECTURE.md\`
-4. \`.agent/agents/prompt-engineer.md\`
+3. \`.agent/VOICE-DESIGN.md\`
+4. \`.agent/ARCHITECTURE.md\`
+5. \`.agent/agents/prompt-engineer.md\`
 5. \`.claude/CLAUDE.md\`
 6. relevant files under \`.claude/rules/\`
 
@@ -242,6 +251,7 @@ Before generating ANY prompts, read skills from \`.agent/skills/\`:
 - Model: \`${config.model}\` (${getModelDisplayName(config.model)})
 - Kling preset: \`${config.klingPreset}\`
 - Create \`${config.outputDir}/project-info.md\`, \`${config.outputDir}/shot-plan.json\`, and \`${config.outputDir}/_index.md\`
+- Keep top-level \`voiceCast\` in \`${config.outputDir}/shot-plan.json\`
 - Write \`${config.outputDir}/shots/SHOTNN.md\` per shot; coverage stays in the same file
 - Refresh \`${config.outputDir}/reports/SAFETY-REPORT.md\` and \`${config.outputDir}/reports/DELIVERY-REPORT.md\` before \`/finish\`
 
@@ -254,11 +264,12 @@ Before generating ANY prompts, read skills from \`.agent/skills/\`:
 6. **Music:** NONE by default.
 7. **Avoid line:** MANDATORY on every prompt (image, video, coverage).
 8. **Coverage:** 2-3 sub-shots within same SHOTNN.md file, min 70 words each.
-9. **ILK/İLK FRAME:** Always include a fenced code block, even when chained.
-10. **Quality Floor:** ILK >= 80, SON >= 80, VIDEO >= 120, coverage >= 70 words.
-11. **Specificity Floor:** lens/framing, lighting, and foreground/midground/background action are mandatory.
-12. **Spatial Realism Floor:** eyeline target, plane map, shared light source, and contact/depth cues are mandatory when relational staging matters.
-13. **ONE FILE PER SHOT:** No separate coverage files.
+9. **Voice Design:** keep project-level \`voiceCast\` in \`${config.outputDir}/shot-plan.json\` and per-shot \`Audio Plan\` in each VIDEO section.
+10. **ILK/İLK FRAME:** Always include a fenced code block, even when chained.
+11. **Quality Floor:** ILK >= 80, SON >= 80, VIDEO >= 120, coverage >= 70 words.
+12. **Specificity Floor:** lens/framing, lighting, and foreground/midground/background action are mandatory.
+13. **Spatial Realism Floor:** eyeline target, plane map, shared light source, and contact/depth cues are mandatory when relational staging matters.
+14. **ONE FILE PER SHOT:** No separate coverage files.
 
 ## Workflows
 | Command | Workflow |
@@ -283,8 +294,9 @@ This workspace keeps high-level policy in \`CLAUDE.md\` and operational detail i
 ## Before Any Generation or Repair
 1. Read \`.agent/model-profile.md\`
 2. Read \`.agent/MASTER.md\`
-3. Read \`.agent/agents/prompt-engineer.md\`
-4. Read \`.agent/workflows/generate.md\` or the requested workflow
+3. Read \`.agent/VOICE-DESIGN.md\`
+4. Read \`.agent/agents/prompt-engineer.md\`
+5. Read \`.agent/workflows/generate.md\` or the requested workflow
 5. Apply AUTO-ANONYMOUS and AUTO-SAFETY before drafting
 6. Prefer the active/selected markdown file as scenario source; fallback is \`${config.scenarioHint}\`
 7. Do not mark work complete while any required report is missing or fail
@@ -293,6 +305,8 @@ This workspace keeps high-level policy in \`CLAUDE.md\` and operational detail i
 - Write only inside \`${config.outputDir}\`
 - Keep one file per shot: \`${config.outputDir}/shots/SHOTNN.md\`
 - Maintain \`${config.outputDir}/shot-plan.json\` dialogue naming policy
+- Maintain \`${config.outputDir}/shot-plan.json\` top-level \`voiceCast\`
+- Keep \`Audio Plan\` blocks aligned to \`voiceCast\`
 - Keep \`ILK/İLK FRAME\` in a fenced code block even when chained
 - Quality floor and specificity floor are hard gates, not suggestions
 - Apply \`.agent/skills/spatial-blocking/SKILL.md\` whenever eyeline, compositing, or depth realism is critical
@@ -313,13 +327,16 @@ Use the Film-Kit core runtime.
 ## Read First
 1. \`.agent/model-profile.md\`
 2. \`.agent/MASTER.md\`
-3. \`.agent/agents/prompt-engineer.md\`
-4. \`.agent/workflows/generate.md\`
-5. \`.claude/rules/output-contract.md\`
+3. \`.agent/VOICE-DESIGN.md\`
+4. \`.agent/agents/prompt-engineer.md\`
+5. \`.agent/workflows/generate.md\`
+6. \`.claude/rules/output-contract.md\`
 
 ## Responsibilities
 - draft and repair shot files under \`${config.outputDir}/shots/\`
 - apply \`${config.outputDir}/shot-plan.json\` dialogue naming policy
+- maintain top-level \`voiceCast\` inside \`${config.outputDir}/shot-plan.json\`
+- keep \`Audio Plan\` blocks valid against \`voiceCast\`
 - enforce AUTO-ANONYMOUS, AUTO-SAFETY, chaining, and coverage contracts
 - enforce quality floor: ILK >= 80, SON >= 80, VIDEO >= 120, coverage >= 70
 - enforce specificity floor: lens/framing, lighting, and foreground/midground/background action
@@ -349,6 +366,7 @@ If using the native Claude subagent, read \`.claude/agents/prompt-engineer.md\`
    - Determine shot count based on action beats
    - Create \`${config.outputDir}/project-info.md\`
    - Create \`${config.outputDir}/shot-plan.json\`
+   - Add top-level \`voiceCast\` before writing speaking shots
 
 2. **Batch Strategy:**
    - 1-10 shots → Generate all at once
@@ -358,6 +376,7 @@ If using the native Claude subagent, read \`.claude/agents/prompt-engineer.md\`
 3. **Per Shot (SINGLE FILE: SHOTNN.md):**
    - Analyze scene type (Dialogue / Action / Emotional / Establishing)
    - Generate main shot (İLK FRAME + SON FRAME + VİDEO)
+   - Add machine-readable \`Audio Plan\` before every VIDEO section
    - Keep İLK FRAME as fenced code block even when chained
    - Enforce hard quality floor: ILK >= 80, SON >= 80, VIDEO >= 120, coverage >= 70
    - Enforce specificity floor: lens/framing + lighting + foreground/midground/background action
@@ -386,8 +405,9 @@ function buildClaudeRuleOutputContract(config) {
 
 ## Required Files
 - \`${config.outputDir}/project-info.md\` — Characters, settings, emotional arc mapping, tension levels
-- \`${config.outputDir}/shot-plan.json\` — Name policy, shot plan, and validation contract
+- \`${config.outputDir}/shot-plan.json\` — Name policy, shot plan, validation contract, and top-level \`voiceCast\`
 - \`.agent/model-profile.md\` — Active model constraints and presets
+- \`.agent/VOICE-DESIGN.md\` — Voice identity and shot audio contract
 - \`${config.outputDir}/_index.md\` — Shot tracking with chain & status
 - \`${config.outputDir}/shots/SHOT01.md ... SHOTNN.md\` — Individual shot files (one file per shot)
 - \`${config.outputDir}/reports/SAFETY-REPORT.md\` — Safety gate report
@@ -403,7 +423,7 @@ The first sentence carries the most weight. Every prompt MUST follow this order:
 3. [Action] Micro-behavior, acting cue, physical gesture
 4. [Camera + Lens] "85mm f/2.0, shallow DOF, static with handheld micro-movement"
 5. [Light + Atmosphere] "Warm oil lamp key light from screen-left, deep shadows"
-6. [Audio direction block] (VIDEO prompts only)
+6. [Audio Plan + Audio direction block] (VIDEO prompts only)
 7. [Avoid line] (EVERY prompt — MANDATORY)
 \`\`\`
 
@@ -413,6 +433,7 @@ The first sentence carries the most weight. Every prompt MUST follow this order:
 - Kling preset: \`${config.klingPreset}\`
 - Kling transition mode: Start+End (when model is kling-3.0)
 - Motion timeline: first → then → finally (when model is kling-3.0)
+- Kling multi-shot mode: single-transition by default; custom storyboard only for 2-3 meaningful phases (when model is kling-3.0)
 
 ## Shot File Format (SHOTNN.md) — SINGLE FILE, ALL INCLUSIVE
 
@@ -447,6 +468,12 @@ Avoid: blurry, low-res, noise, distorted faces, bad anatomy, extra limbs/fingers
 plastic skin, waxy skin, on-screen text, watermark, logo, cartoon style, CGI look.
 \\\`\\\`\\\`
 
+### AUDIO PLAN
+
+\\\`\\\`\\\`json
+[Machine-readable audioPlan JSON block aligned with shot-plan.json -> voiceCast]
+\\\`\\\`\\\`
+
 ### VİDEO
 
 \\\`\\\`\\\`
@@ -494,6 +521,10 @@ Audio direction:
 Avoid: distorted faces, morphing, blurry, flickering, unnatural motion, on-screen text.
 \\\`\\\`\\\`
 
+\\\`\\\`\\\`json
+[Coverage audioPlan JSON block - voice binding only when coverage contains speech]
+\\\`\\\`\\\`
+
 ### SHOT[NN]B — [Type] | [Duration]s | [Icon] [Label]
 [Same format as A]
 
@@ -545,6 +576,8 @@ Total: 1 main + [N] coverage = [N+1] production shots
 ### Name Rule
 - Visual prompts and non-dialogue fields: no real names
 - Dialogue transcript naming follows \`shot-plan.json.dialogue_name_policy\`
+- Every speaking shot must resolve \`activeSpeakerKey\` to \`shot-plan.json.voiceCast\`
+- Keep one active speaker per shot
 
 ### 30° Rule
 Coverage camera angles MUST differ from main shot by at least 30°.
@@ -590,6 +623,7 @@ function buildCopilotInstructions(config) {
 ### System Entry Point
 Read \`.agent/MASTER.md\` for the complete production ruleset.
 Read \`.agent/ARCHITECTURE.md\` for system overview.
+Read \`.agent/VOICE-DESIGN.md\` for voice identity and shot audio contracts.
 Read \`.agent/model-profile.md\` for active model rules.
 
 ### Skill Loading Protocol (MANDATORY)
@@ -610,7 +644,8 @@ Before generating ANY prompts, read skills from \`.agent/skills/\`:
 4. Create index: \`${config.outputDir}/_index.md\`
 5. Create project info: \`${config.outputDir}/project-info.md\`
 6. Create plan: \`${config.outputDir}/shot-plan.json\`
-7. Write reports: \`${config.outputDir}/reports/SAFETY-REPORT.md\`, \`${config.outputDir}/reports/DELIVERY-REPORT.md\`
+7. Keep top-level \`voiceCast\` in the plan and \`Audio Plan\` in speaking VIDEO sections
+8. Write reports: \`${config.outputDir}/reports/SAFETY-REPORT.md\`, \`${config.outputDir}/reports/DELIVERY-REPORT.md\`
 
 ### Critical Rules
 - **AUTO-ANONYMOUS:** Replace ALL real names with physical descriptions
@@ -621,6 +656,7 @@ Before generating ANY prompts, read skills from \`.agent/skills/\`:
 - **Spatial Realism:** eyeline targets, shared light, depth scale, and anti-cutout staging must agree when subjects share frame
 - **Avoid Line:** MANDATORY on every prompt
 - **Music:** NONE by default
+- **Voice Design:** keep \`voiceCast\` in \`${config.outputDir}/shot-plan.json\` and \`Audio Plan\` in speaking VIDEO sections
 - **Duration:** 8s default, slow burn pacing
 - **Language:** Prompts in English, dialogue preserved
 - **ILK/İLK FRAME:** keep fenced code block even when chained
@@ -650,9 +686,10 @@ When request is /generate, follow the Film-Kit Hollywood production system:
 3. Load required skills from \`.agent/skills/\`
 4. Transform scenario into production shot package at \`${config.outputDir}\`
 5. Generate: project-info.md, shot-plan.json, _index.md, shots/SHOT01.md..SHOTNN.md
-6. Each SHOTNN.md: İLK FRAME + SON FRAME + VİDEO + 2-3 Coverage (ALL IN ONE FILE)
-7. Enforce: auto-anonymous, dialogue name policy, auto-safety, frame chaining, avoid lines
-8. Write reports to \`${config.outputDir}/reports/\` before /finish
+6. Keep top-level \`voiceCast\` in shot-plan.json
+7. Each SHOTNN.md: İLK FRAME + SON FRAME + AUDIO PLAN + VİDEO + 2-3 Coverage (ALL IN ONE FILE)
+8. Enforce: auto-anonymous, dialogue name policy, auto-safety, frame chaining, avoid lines
+9. Write reports to \`${config.outputDir}/reports/\` before /finish
 `;
 }
 /* ---------- ANTIGRAVITY ---------- */
@@ -667,6 +704,7 @@ description: Hollywood-standard cinematic prompt engineering with model-aware pr
 ## System Architecture
 This skill is part of the Film-Kit prompt engineering system.
 Read \`.agent/MASTER.md\` for the complete production ruleset (690+ rules).
+Read \`.agent/VOICE-DESIGN.md\` for project-level \`voiceCast\` and shot-level \`audioPlan\`.
 Read \`.agent/model-profile.md\` first for active model constraints.
 
 ## Skill Loading Protocol
@@ -700,6 +738,7 @@ Before generating ANY prompts, read these skills:
 - Shot index: \`${config.outputDir}/_index.md\`
 - Project info: \`${config.outputDir}/project-info.md\`
 - Plan: \`${config.outputDir}/shot-plan.json\`
+- Voice contract: top-level \`voiceCast\` in \`${config.outputDir}/shot-plan.json\`
 - Reports: \`${config.outputDir}/reports/SAFETY-REPORT.md\`, \`${config.outputDir}/reports/DELIVERY-REPORT.md\`
 
 ## Critical Rules
@@ -709,12 +748,13 @@ Before generating ANY prompts, read these skills:
 4. **Frame Chaining:** Last frame of SHOT[N] becomes first frame of SHOT[N+1]
 5. **Coverage Mandatory:** 2-3 sub-shots per main shot (in same file, min 60 words each)
 6. **Avoid Line:** MANDATORY on every prompt (image + video + coverage)
-7. **Music: NONE** by default
-8. **Ultra Realism** default visual mode
-9. **8s duration** default, slow burn pacing
-10. **ILK/İLK FRAME:** always keep fenced code block
-11. **ONE FILE PER SHOT:** SHOTNN.md contains main shot + all coverage
-12. **Relational Realism:** preserve eyeline targets, shared light, depth scale, and anti-cutout staging when multiple subjects share frame
+7. **Voice Design:** keep \`voiceCast\` in \`${config.outputDir}/shot-plan.json\` and \`Audio Plan\` in every speaking VIDEO section
+8. **Music: NONE** by default
+9. **Ultra Realism** default visual mode
+10. **8s duration** default, slow burn pacing
+11. **ILK/İLK FRAME:** always keep fenced code block
+12. **ONE FILE PER SHOT:** SHOTNN.md contains main shot + all coverage
+13. **Relational Realism:** preserve eyeline targets, shared light, depth scale, and anti-cutout staging when multiple subjects share frame
 
 ## Quality Floor (Hard Gate)
 Reject and regenerate any shot that fails:
@@ -843,6 +883,16 @@ Limit the amount of change per clip for realism:
 
 > If end frame is too different, the model must "invent miracles" → plastic/melting/warp risk rises.
 
+### Kling Storyboard / Multi-Shot Decision Tree
+
+The official Kling app surface exposes custom storyboard-like shot prompting and optional end-frame anchoring.
+Use it as a selective upgrade, not a default:
+- **single-transition**: default mode for one clean action arc, reveal, or emotional turn
+- **custom-storyboard**: only when one clip truly needs **2-3 editorially distinct phases**
+- **hard cap in this toolkit:** 3 custom storyboard shots per video generation
+- **never storyboard micro-beats:** one glance, one finger move, one tiny prop touch, one breath shift
+- if the beat needs 4+ phases, split into chained videos instead of overloading one Kling prompt
+
 ### Golden Prompt Skeleton (Start+End)
 
 The prompt's job is to tell the model **how to generate the in-between frames**:
@@ -984,26 +1034,22 @@ When using a reference image as the start frame, the model extracts lighting and
 
 ### Multi-Shot Protocol (Tek Üretimde Çoklu Çekim)
 
-Kling 3.0 can manage **2 to 6 shots** in a single 15-second generation.
+Treat Kling multi-shot as **storyboarded internal progression**, not as a license to over-cut.
 
 **Rules:**
-1. **Shot Prefix:** Each shot MUST begin with \`Shot X,\` (e.g., \`Shot 1,\`, \`Shot 2,\`)
-2. **Character Continuity:** Repeat physical descriptions at the start of each shot, or use **Element Binding** (see below)
-3. **Time Distribution:** 15 seconds must be logically divided (e.g., Shot 1: 3s, Shot 2: 7s, Shot 3: 5s)
-4. **Maximum:** 6 shots per generation; for more, use separate generations with chaining
-
-Official examples favor short, concrete camera-and-action openings. Start with the movement path or subject action, then lock the stable identity/background constraints.
-
-**Example Multi-Shot Prompt:**
-\`\`\`
-Shot 1, Close-up of a bearded man in his 40s, wearing a dusty military uniform. He stares ahead, jaw clenched. Tripod-locked, 85mm, shallow DOF. (3s)
-
-Shot 2, Medium shot, same bearded man turns toward a younger soldier standing behind him. Slow pan right to reveal the young soldier. Natural handheld micro-sway. (5s)
-
-Shot 3, Over-the-shoulder from the younger soldier's POV, the bearded man speaks directly to camera. Gentle push-in. (4s)
-
-Shot 4, Wide shot establishing the artillery position, both soldiers visible. Crane rise. (3s)
-\`\`\`
+1. Default to **single-transition** whenever one camera path can carry the beat.
+2. Use custom storyboard only for **2-3 distinct internal phases** with different framing/action purpose.
+3. Keep each storyboard phase short, clear, and concrete: what changed, what stayed locked, why the cut exists.
+4. Hard cap for this toolkit: **3 storyboard shots per generation** in app.kling-oriented workflows.
+5. If the sequence wants 4+ phases, split into multiple chained generations.
+6. Never assign separate storyboard shots to micro-actions that would read better inside one stronger shot.
+
+**Preferred storyboard jobs:**
+- Shot 1: establish or setup
+- Shot 2: main action / reveal / shift
+- Shot 3: reaction / settle / handoff
+
+Official Kling surfaces favor short, concrete shot descriptions. Start from the action path or camera intention, then lock continuity anchors (identity, wardrobe, background geometry, stable light).
 
 ### Element Binding (Öğe Bağlama — Karakter/Nesne Tutarlılığı)
 

package/content/ARCHITECTURE.md

@@ -20,6 +20,7 @@ Modular system consisting of:
 .agent/
 ├── ARCHITECTURE.md          # This file
 ├── MASTER.md                # Main rules (entry point for AI tools)
+├── VOICE-DESIGN.md          # Voice identity + audioPlan contract
 ├── model-profile.md         # Active model rules (runtime generated)
 ├── agents/
 │   └── prompt-engineer.md   # Primary agent
@@ -60,7 +61,7 @@ Modular system consisting of:
 | `coverage-system` | **Mandatory coverage shots** (Reaction, OTS, Insert, Cutaway, ECU, Wide) + L-cut/J-cut + 30° kuralı + **180° kuralı** + eyeline match + matching action + multi-character blocking |
 | `spatial-blocking` | **Relational realism**: eyeline targeting, plane mapping, body orientation, shared lighting, depth/scale integration, anti-cutout / anti-miniature cues |
 | `visual-modes` | **Ultra Realism** default, stylization triggers, anti-AI artifact rules + **renk sürekliliği** + magic hour + flashback/rüya görsel ayrımı |
-| `audio-design` | **Sound design** rules, voice realism, SFX, ambience, audio direction block + diegetic/non-diegetic ses ayrımı |
+| `audio-design` | **Sound design** rules, voice realism, project-level `voiceCast`, shot-level `audioPlan`, audio direction block + diegetic/non-diegetic ses ayrımı |
 | `prompt-structure` | Image/video prompt templates, camera vocabulary, seed parameter, prompt rewriter, **re-take strategy**, coverage prompt yazım standartları (≥60 kelime) |
 
 ---
@@ -84,6 +85,8 @@ User Scenario → Agent Activated → Read model-profile → Load Required Skill
                                        ↓
                             .agent/model-profile.md (ALWAYS FIRST)
                                        ↓
+                            .agent/VOICE-DESIGN.md (when dialogue / narrator exists)
+                                       ↓
                             safety-compliance (ALWAYS)
                             reference-locking (if refs provided)
                             frame-chaining (ALWAYS)

package/content/MASTER.md

@@ -131,9 +131,23 @@ User: "Subay askerlere bağırır ve ateş emri verir."
   2. "Ateş!" diye bağırır, damarları çıkar. (Mid-shot)
   3. Askerlerin parmağı tetiğe gider. (Detail shot)
 
-### 🎵 Music Policy
-**DEFAULT:** `Music: NONE` (Always).
-User must explicitly request music to include it. Otherwise, rely on SFX and Ambience.
+### 🎵 Music Policy
+**DEFAULT:** `Music: NONE` (Always).
+User must explicitly request music to include it. Otherwise, rely on SFX and Ambience.
+
+### Voice Design Contract
+
+When dialogue or narration exists, audio is mandatory at three synchronized layers:
+
+1. project-level `voiceCast`
+2. shot-level `audioPlan`
+3. prompt-level `Audio direction:`
+
+Project-level voice identity is stable.
+Shot-level performance is variable.
+
+Use `.agent/VOICE-DESIGN.md` whenever a speaker, narrator, or voiceover is involved.
+Do not redesign a speaker voice per shot. Reuse the same `speakerKey` and voice identity across the whole project.
 
 ### 📸 Reference Image Enforcement
 
@@ -160,7 +174,7 @@ When user provides references, EVERY prompt MUST:
 3. [Aksiyon] — Ne yapıyor, mikro-davranış, oyunculuk ipucu
 4. [Kamera + Lens] — "85mm f/2.0, shallow DOF, static with handheld micro-movement"
 5. [Işık + Atmosfer] — "Warm oil lamp key light from screen-left, deep shadows"
-6. [Audio direction block] — (sadece video prompt'larında)
+6. [Audio Plan + Audio direction block] — (sadece video prompt'larında)
 7. [Avoid line] — (HER prompt'ta zorunlu)
 ```
 
@@ -186,11 +200,20 @@ Avoid: blurry, low-res, noise, distorted faces, bad anatomy, extra limbs/fingers
 - `--audio_mode` or `--audio_prompt` (audio is for VIDEO only)
 - Any `--parameter` flags
 
-#### VIDEO PROMPTS (VİDEO, Coverage Video)
-
-**MUST INCLUDE FULL AUDIO DIRECTION BLOCK:**
-```
-Audio direction:
+#### VIDEO PROMPTS (VİDEO, Coverage Video)
+
+**VOICE DESIGN CONTRACT (MANDATORY WHEN VOICE EXISTS):**
+- Project output must include top-level `voiceCast`
+- Every VIDEO section must include a machine-readable `Audio Plan` block
+- If `Type` is `Dialogue`, `Voiceover`, or `Mixed`, `activeSpeakerKey` must bind to `voiceCast`
+- `voiceIdentityPrompt` is character-level
+- `performanceNote` is shot-level
+- Keep one active speaker per shot
+- Split reply dialogue across multiple shots when needed
+
+**MUST INCLUDE FULL AUDIO DIRECTION BLOCK:**
+```
+Audio direction:
 - Language: [TURKISH/ENGLISH/etc.]
 - Type: [Dialogue/SFX/Ambience/Mixed]
 - Dialogue transcript: "[Exact lines in original language]" or NONE
@@ -530,11 +553,13 @@ Avoid: blurry, low-res, noise, distorted faces, bad anatomy, extra limbs/fingers
 **VİDEO:**
 
 ```
-[Complete video prompt — MIN 80 words, MAX 120 words]
-[Action + camera movement + acting cues + atmosphere]
-
-Audio direction:
-- Language: [LANGUAGE]
+[Complete video prompt — MIN 80 words, MAX 120 words]
+[Action + camera movement + acting cues + atmosphere]
+
+[Audio Plan JSON block aligned with .agent/VOICE-DESIGN.md]
+
+Audio direction:
+- Language: [LANGUAGE]
 - Type: [Dialogue/SFX/Ambience]
 - Dialogue transcript: [Lines or NONE]
 - SFX: [Effects list]
@@ -613,11 +638,15 @@ Before outputting, validate EVERY shot. **Bu kontrol otomatiktir, kullanıcı ha
 - [ ] **HER coverage image** → Avoid satırı var mı? ❗ Yoksa EKLE
 - [ ] **HER coverage video** → Avoid satırı var mı? ❗ Yoksa EKLE
 
-### 4️⃣ Prompt Uzunluk Kontrolü
-- [ ] Ana shot image prompt ≥ 60 kelime mi?
-- [ ] Ana shot video prompt ≥ 80 kelime mi?
-- [ ] **Coverage prompt ≥ 60 kelime mi?** ❗ Kısa coverage YASAK
-- [ ] Audio direction block tam mı? (Language, Type, Dialogue, SFX, Ambience, Music, Mix)
+### 4️⃣ Prompt Uzunluk Kontrolü
+- [ ] Ana shot image prompt ≥ 60 kelime mi?
+- [ ] Ana shot video prompt ≥ 80 kelime mi?
+- [ ] **Coverage prompt ≥ 60 kelime mi?** ❗ Kısa coverage YASAK
+- [ ] Audio direction block tam mı? (Language, Type, Dialogue, SFX, Ambience, Music, Mix)
+- [ ] `voiceCast` mevcut mu ve her konusan speaker icin `speakerKey` tekil mi?
+- [ ] Her VIDEO bolumunde machine-readable `Audio Plan` var mi?
+- [ ] Dialogue / Voiceover shot'larinda `activeSpeakerKey` -> `voiceCast` baglantisi gecerli mi?
+- [ ] Bir shot'ta tek aktif speaker kurali korunuyor mu?
 
 ### 5️⃣ Türkçe Özet
 - [ ] Her shot'un 🇹🇷 Türkçe özet satırı var mı?

package/content/RULES.md

@@ -114,10 +114,11 @@ When references provided:
 
 | Kural | Açıklama |
 |-------|----------|
-| HER image prompt | Avoid satırı ile bitmeli |
-| HER video prompt | Audio direction block + Avoid satırı ile bitmeli |
-| HER coverage prompt | ≥ 60 kelime + Avoid satırı ile bitmeli |
-| Türkçe özet | Her shot ve coverage için 🇹🇷 özet satırı |
+| HER image prompt | Avoid satırı ile bitmeli |
+| HER video prompt | Audio direction block + Avoid satırı ile bitmeli |
+| HER coverage prompt | ≥ 60 kelime + Avoid satırı ile bitmeli |
+| Türkçe özet | Her shot ve coverage için 🇹🇷 özet satırı |
+| Voice Design | `voiceCast` proje seviyesi, `audioPlan` shot seviyesi, `Audio direction` parser seviyesi |
 
 ### 📷 Coverage System (MANDATORY)
 
@@ -134,10 +135,20 @@ When references provided:
 
 **Sub-shot naming:** SHOT05A, SHOT05B, SHOT05C (main shot + alphabetical suffix)
 
-### 🎵 Music Policy (BANNED)
-
-**DEFAULT:** `Music: NONE` (Always).
-User must explicitly request music. Otherwise use SFX/Ambience only.
+### 🎵 Music Policy (BANNED)
+
+**DEFAULT:** `Music: NONE` (Always).
+User must explicitly request music. Otherwise use SFX/Ambience only.
+
+### Voice Design Policy
+
+If dialogue or narration exists:
+
+- keep a project-level `voiceCast`
+- keep a shot-level `audioPlan`
+- keep the prompt-level `Audio direction:` block
+- keep one active speaker per shot
+- reuse the same `speakerKey` and voice identity across all shots
 
 ### 🎭 Dramaturji & Oyunculuk (Diyalog Sahneleri)
 

package/content/VOICE-DESIGN.md

@@ -0,0 +1,248 @@
+# Voice Design Output Contract
+
+This runtime must support voice-design style audio generation where voice identity is created once per speaker and then reused across shots.
+
+## Three-Layer Audio Architecture
+
+Every scenario output must keep these layers aligned:
+
+1. `voiceCast`
+   Project-level or scenario-level voice identity package.
+2. `audioPlan`
+   Shot-level performance, transcript, SFX, ambience, and mix metadata.
+3. `Audio direction:`
+   Human-readable prompt block kept for backward compatibility with existing parsers.
+
+## Film-Kit Storage Contract
+
+- Single-agent projects store project-level voice identity in `$OUTPUT_DIR/shot-plan.json`.
+- Multi-agent projects store project-level voice identity in `$OUTPUT_DIR/team-plan.json`.
+- Every `SHOTNN.md` file must include a machine-readable `Audio Plan` JSON block for each VIDEO section.
+- The `Audio direction:` prompt block must mirror the same `audioPlan` and remain present in the video prompt.
+
+## Core Principle
+
+Voice identity is character-level.
+Performance is shot-level.
+
+Do not redesign a character voice on every shot.
+Design the voice once, bind it to a stable `speakerKey`, and reuse that identity across the entire project.
+
+## Project-Level `voiceCast`
+
+Create one `voiceCast` entry for every speaking character or narrator.
+
+Required fields:
+
+- `speaker`
+- `speakerKey`
+- `role`
+- `language`
+- `voiceIdentityPrompt`
+- `voicePerformanceBase`
+- `preferredProvider`
+- `preferredModel`
+- `saveToLibrary`
+
+Recommended fields:
+
+- `genderHint`
+- `ageRange`
+- `referenceAudioStrategy`
+- `referenceAudioNeeded`
+- `shouldGenerateVoice`
+- `voiceDesignRecommended`
+- `voiceDesignPriority`
+- `referenceAudioSuggested`
+- `referenceAudioDescription`
+- `notes`
+
+Example:
+
+```json
+{
+  "speaker": "Kemal",
+  "speakerKey": "kemal",
+  "role": "character",
+  "language": "turkish",
+  "genderHint": "male",
+  "ageRange": "35-45",
+  "voiceIdentityPrompt": "A realistic Turkish male voice in his late thirties to early forties. Low-mid register, restrained authority, subtle fatigue, natural diction, intimate close-mic realism, no theatricality, highly believable human texture.",
+  "voicePerformanceBase": "controlled, interior, understated",
+  "referenceAudioStrategy": "optional",
+  "referenceAudioNeeded": false,
+  "shouldGenerateVoice": true,
+  "preferredProvider": "elevenlabs",
+  "preferredModel": "eleven_ttv_v3",
+  "saveToLibrary": true,
+  "voiceDesignRecommended": true,
+  "voiceDesignPriority": "high",
+  "referenceAudioSuggested": false,
+  "referenceAudioDescription": null,
+  "notes": "Use the same voice identity across all shots. Performance may change per shot, identity must not."
+}
+```
+
+## `speakerKey` Rule
+
+The same speaker must use the same `speakerKey` everywhere.
+If a dialogue shot references a `speakerKey` that does not exist in `voiceCast`, treat it as a contract failure.
+
+## `voiceIdentityPrompt` Rule
+
+`voiceIdentityPrompt` describes the stable voice identity, not the temporary emotion of one shot.
+
+Include:
+
+- language
+- age range
+- gender feel when useful
+- register and texture
+- diction and articulation
+- energy profile
+- recording feel or mic proximity
+- realism target such as believable, non-theatrical, non-announcer
+
+Do not include:
+
+- exact dialogue lines
+- one-shot emotional spikes
+- scene-specific action like running, shouting, whispering this exact line
+
+## `voicePerformanceBase`
+
+Use `voicePerformanceBase` for the speaker's normal acting baseline.
+This baseline may combine with the shot-level `performanceNote`, but it must not replace voice identity.
+
+## Shot-Level `audioPlan`
+
+Every VIDEO section must carry a machine-readable `Audio Plan` JSON block.
+
+If dialogue or voiceover exists, these fields are mandatory:
+
+- `language`
+- `type`
+- `activeSpeaker`
+- `activeSpeakerKey`
+- `dialogueLines`
+- `dialogueTranscript`
+- `performanceNote`
+
+Recommended support fields:
+
+- `delivery.pace`
+- `delivery.intensity`
+- `delivery.emotion`
+- `delivery.projection`
+- `delivery.stability`
+- `sfx`
+- `ambience`
+- `music`
+- `mixTarget`
+- `hasSubtitles`
+
+Example:
+
+```json
+{
+  "language": "turkish",
+  "type": "dialogue",
+  "activeSpeaker": "Kemal",
+  "activeSpeakerKey": "kemal",
+  "dialogueLines": [
+    {
+      "speaker": "Kemal",
+      "speakerKey": "kemal",
+      "text": "Burayi terk etmemiz lazim."
+    }
+  ],
+  "dialogueTranscript": "Kemal: Burayi terk etmemiz lazim.",
+  "performanceNote": "Low, controlled urgency. Quiet but decisive. No melodrama.",
+  "delivery": {
+    "pace": "measured",
+    "intensity": "medium-low",
+    "emotion": "suppressed urgency",
+    "projection": "intimate",
+    "stability": "steady"
+  },
+  "sfx": [
+    "Ahsap sandalyenin hafif surtunmesi"
+  ],
+  "ambience": [
+    "Kucuk odada dusuk floresan humu"
+  ],
+  "music": null,
+  "mixTarget": {
+    "dialogue": 70,
+    "sfx": 5,
+    "ambience": 25
+  },
+  "hasSubtitles": false
+}
+```
+
+## Single Active Speaker Rule
+
+For realistic lipsync and stable delivery, keep only one active speaker per shot.
+
+- Other characters may appear in frame, but they should listen silently.
+- Back-and-forth dialogue should be split across multiple shots.
+- Coverage shots may have `audioPlan.type` values like `sfx-only` or `ambience-only` and therefore require no voice binding.
+
+## Backward-Compatible Prompt Block
+
+The prompt must still include the human-readable block:
+
+```text
+Audio direction:
+Language: Turkish
+Type: Dialogue
+Dialogue transcript:
+Kemal: Burayi terk etmemiz lazim.
+SFX:
+- Ahsap sandalyenin hafif surtunmesi
+Ambience:
+- Kucuk odada dusuk floresan humu
+Music: NONE
+Mix target: Dialogue 70%, SFX 5%, Ambience 25%
+No on-screen subtitles/captions.
+```
+
+This block must be derived from the same `audioPlan`, not invented separately.
+
+## Default Voice Design Policy
+
+- Character voice stays fixed.
+- Shot performance may change.
+- Language and diction stay fixed.
+- Realism is maximized.
+- Theatricality is minimized.
+- Promotional or announcer tone is forbidden unless explicitly requested.
+- `Music: NONE` is the default.
+- `No on-screen subtitles/captions.` is mandatory unless the user explicitly requests otherwise.
+
+## Reference Audio Guidance
+
+Recommend reference audio when the target voice depends on unusually specific texture:
+
+- elderly or highly specific age character
+- broken, smoky, hoarse, or damaged micro-texture
+- strong local accent or highly specific diction
+- very narrow performance tolerance between whisper and force
+
+If reference audio is recommended, prefer:
+
+- `preferredModel: "eleven_ttv_v3"`
+- `referenceAudioNeeded: true`
+
+## Provider Flow
+
+When a provider supports voice design, the implementation flow is:
+
+1. Use `voiceCast[].voiceIdentityPrompt` to design a preview voice.
+2. Review multiple previews.
+3. Select one preview.
+4. Create a persistent `voice_id`.
+5. Bind that `voice_id` to `speakerKey`.
+6. Reuse the same voice identity across all future shots.
+7. If needed, remix or refine the same voice without breaking identity continuity.

package/content/agents/prompt-engineer.md

@@ -133,13 +133,24 @@ First sentence carries the most weight. Every prompt MUST follow this order:
 3. [Action] Micro-behavior, acting cue, physical gesture
 4. [Camera + Lens] "85mm f/2.0, shallow DOF, static with handheld micro-movement"
 5. [Light + Atmosphere] "Warm oil lamp key light from screen-left, deep shadows"
-6. [Audio direction block] (VIDEO prompts only)
-7. [Avoid line] (EVERY prompt — MANDATORY)
-```
-
-> **Kling adaptation:** When model is kling-3.0, replace step 3 (Action) with motion timeline (`first → then → finally`). Step 6 uses Kling tone markers instead of Audio Direction Block. See `.agent/model-profile.md` for golden prompt skeleton.
-
----
+6. [Audio direction block] (VIDEO prompts only)
+7. [Avoid line] (EVERY prompt — MANDATORY)
+```
+
+> **Kling adaptation:** When model is kling-3.0, replace step 3 (Action) with motion timeline (`first → then → finally`). Step 6 uses Kling tone markers instead of Audio Direction Block. See `.agent/model-profile.md` for golden prompt skeleton.
+
+## VOICE DESIGN CONTRACT
+
+When dialogue, narrator VO, or character speech exists:
+
+- read `.agent/VOICE-DESIGN.md`
+- create and maintain project-level `voiceCast` in `shot-plan.json`
+- reuse the same `speakerKey` across all shots
+- write a machine-readable `Audio Plan` JSON block for every VIDEO section
+- keep `voiceIdentityPrompt` character-level and `performanceNote` shot-level
+- keep one active speaker per shot
+
+---
 
 ## AUDIO DIRECTION (Mandatory for Video)
 
@@ -195,22 +206,29 @@ Before outputting ANY shot:
 - [ ] No active violence/gore?
 
 ### 2. Prompt Structure
-- [ ] Follows 1-7 Flow Order?
-- [ ] Avoid Line present on EVERY prompt (Image + Video + Coverage)?
-- [ ] Audio direction present on EVERY video prompt? (Veo block or Kling inline)
-- [ ] 2-3 Coverage shots included in the same file?
+- [ ] Follows 1-7 Flow Order?
+- [ ] Avoid Line present on EVERY prompt (Image + Video + Coverage)?
+- [ ] Audio direction present on EVERY video prompt? (Veo block or Kling inline)
+- [ ] `shot-plan.json` contains top-level `voiceCast`?
+- [ ] Every speaking shot has an `Audio Plan` block with valid `activeSpeakerKey`?
+- [ ] Each active speaker resolves to a stable `speakerKey` in `voiceCast`?
+- [ ] One active speaker per dialogue shot?
+- [ ] 2-3 Coverage shots included in the same file?
 - [ ] Quality floor passes? (ILK>=80, SON>=80, VIDEO>=120, coverage>=70)
 - [ ] Specificity floor passes? (lens + lighting + FG/MG/BG action)
 - [ ] Spatial realism passes? (eyeline target + plane map + shared light + contact/depth cues)
 - [ ] Model Control block exists? (`Model`, `Preset`, `CFG`, `Transition Mode`)
 
-### 3. Kling-Specific Gates (when model is kling-3.0)
-- [ ] Motion timeline uses `first → then → finally` structure?
-- [ ] "What stays the same" explicitly stated? (identity, background, costume)
-- [ ] Camera movement is simple/safe? (no complex hybrid movements)
-- [ ] Negative prompt includes Kling cleanup set? (warping, rubbery, melted)
-- [ ] Duration matches transformation budget? (5s=1 change, 10s=2-3, 15s=complex)
-- [ ] Start/end frames are in same visual universe? (angle, scale, light, lens)
+### 3. Kling-Specific Gates (when model is kling-3.0)
+- [ ] Motion timeline uses `first → then → finally` structure?
+- [ ] "What stays the same" explicitly stated? (identity, background, costume)
+- [ ] Camera movement is simple/safe? (no complex hybrid movements)
+- [ ] Negative prompt includes Kling cleanup set? (warping, rubbery, melted)
+- [ ] Duration matches transformation budget? (5s=1 change, 10s=2-3, 15s=complex)
+- [ ] Start/end frames are in same visual universe? (angle, scale, light, lens)
+- [ ] Used `single-transition` by default instead of forcing storyboard complexity?
+- [ ] If custom storyboard is used, is it capped at 3 meaningful stages?
+- [ ] Did each storyboard stage earn a distinct editorial job instead of representing micro-gestures?
 
 ---
 
@@ -241,15 +259,20 @@ For EACH shot, output exactly one file (`SHOTNN.md`) containing Main Shot + Cove
 [If CHAINED: include "Use SHOT[prev]_END as exact first frame" inside code block]
 [If FIRST/BREAK: image prompt in same code block]
 
-### SON FRAME (SHOTNN_END)
-```
-[Image Prompt (Flow Order + Avoid Line)]
-avoid: blurry, low-res, text, watermark, bad anatomy, distorted face ...
-```
-
-### VİDEO
-```
-[Video Prompt (Flow Order)]
+### SON FRAME (SHOTNN_END)
+```
+[Image Prompt (Flow Order + Avoid Line)]
+avoid: blurry, low-res, text, watermark, bad anatomy, distorted face ...
+```
+
+### AUDIO PLAN
+```json
+[Machine-readable audioPlan JSON block aligned with shot-plan.json -> voiceCast]
+```
+
+### VİDEO
+```
+[Video Prompt (Flow Order)]
 
 Audio direction:
 - Language: ...

package/content/skills/audio-design/SKILL.md

@@ -3,12 +3,136 @@ name: audio-design
 description: Sound design rules for model-aware profiles (Veo 3.1 / Kling 3.0). Voice realism, environmental sounds, SFX, ambience, and audio direction block formatting. Includes Kling native audio tonlama and anti-synthetic audio guidelines.
 ---
 
-# Audio Design System
-
-> **Philosophy:** Sound is half the experience. Audio must be as real as visuals.
-> **Core Principle:** Every sound recorded-quality. No synthetic giveaways.
-
----
+# Audio Design System
+
+> **Philosophy:** Sound is half the experience. Audio must be as real as visuals.
+> **Core Principle:** Every sound recorded-quality. No synthetic giveaways.
+
+---
+
+## Voice Design Contract (MANDATORY)
+
+When a scenario includes spoken dialogue, narrator VO, or any reusable character voice, use a three-layer contract:
+
+1. `voiceCast`
+   Stable project-level voice identity package.
+2. `audioPlan`
+   Shot-level performance and mix metadata.
+3. `Audio direction:`
+   Human-readable prompt block for backward compatibility.
+
+Use `.agent/VOICE-DESIGN.md` as the authoritative schema reference.
+
+### Project-Level `voiceCast`
+
+Every speaking character or narrator must have one `voiceCast` entry.
+
+Required fields:
+
+- `speaker`
+- `speakerKey`
+- `role`
+- `language`
+- `voiceIdentityPrompt`
+- `voicePerformanceBase`
+- `preferredProvider`
+- `preferredModel`
+- `saveToLibrary`
+
+Rules:
+
+- `speakerKey` must stay stable across the whole project.
+- `voiceIdentityPrompt` describes the permanent voice identity, never a one-shot emotional spike.
+- `voicePerformanceBase` describes the speaker's normal acting baseline.
+- If a dialogue shot has no matching `voiceCast` entry, the output is invalid.
+
+Example:
+
+```json
+{
+  "speaker": "Kemal",
+  "speakerKey": "kemal",
+  "role": "character",
+  "language": "turkish",
+  "voiceIdentityPrompt": "A realistic Turkish male voice in his late thirties to early forties. Low-mid register, restrained authority, subtle fatigue, natural diction, intimate close-mic realism, no theatricality, highly believable human texture.",
+  "voicePerformanceBase": "controlled, interior, understated",
+  "preferredProvider": "elevenlabs",
+  "preferredModel": "eleven_ttv_v3",
+  "saveToLibrary": true
+}
+```
+
+### Shot-Level `audioPlan`
+
+Every VIDEO block must include a machine-readable `Audio Plan` JSON section that maps to the project-level voice identity.
+
+If the shot contains dialogue or voiceover, these fields are mandatory:
+
+- `language`
+- `type`
+- `activeSpeaker`
+- `activeSpeakerKey`
+- `dialogueLines`
+- `dialogueTranscript`
+- `performanceNote`
+
+Recommended support fields:
+
+- `delivery.pace`
+- `delivery.intensity`
+- `delivery.emotion`
+- `delivery.projection`
+- `delivery.stability`
+- `sfx`
+- `ambience`
+- `music`
+- `mixTarget`
+- `hasSubtitles`
+
+### Single Active Speaker Rule
+
+Keep only one active speaker per shot.
+
+- Reaction shots should usually be silent or ambience-only.
+- Back-and-forth dialogue should be split across multiple shots.
+- OTS coverage may carry one voice, but do not ask the model to lipsync two people at once.
+
+### Voice Identity vs Performance
+
+Keep this distinction strict:
+
+- `voiceIdentityPrompt` = who the speaker is across the whole film
+- `performanceNote` = how that speaker performs in this shot
+
+Correct identity cues:
+
+- `low-mid register`
+- `natural Turkish diction`
+- `restrained authority`
+- `breathy but realistic`
+- `non-theatrical`
+- `highly believable real-world tone`
+
+Incorrect identity cues:
+
+- exact dialogue lines
+- `angrily says this line`
+- `whispering while running`
+- temporary scene emotion only
+
+### Provider Flow
+
+Voice Design style providers should follow this order:
+
+1. design from `voiceCast[].voiceIdentityPrompt`
+2. review previews
+3. choose one
+4. create persistent voice
+5. bind voice to `speakerKey`
+6. reuse across all shots
+7. remix only as a refinement, not as a new character identity
+
+---
 
 ## 🎯 Audio Realism Baseline
 
@@ -197,19 +321,22 @@ User: "Kalkın aslanlar! Düşman zırhlıları top menziline giriyorlar!"
 ✅ RIGHT: Dialogue transcript: "Kalkın aslanlar! Düşman zırhlıları top menziline giriyorlar!"
 ```
 
-### When User Provides Dialogue
-
-- Include VERBATIM in audio transcript
-- Preserve original language (Turkish stays Turkish, English stays English)
-- Note emotional delivery required
+### When User Provides Dialogue
+
+- Include VERBATIM in audio transcript
+- Preserve original language (Turkish stays Turkish, English stays English)
+- Note emotional delivery required
+- Mirror the same transcript in `audioPlan.dialogueLines`
+- Bind the speaking line to `activeSpeakerKey`
 
 ### 🗣️ SPEAKER ISOLATION RULE (Prevent Mixed Dialogue)
 
 **Problem:** If two people are in the frame and both speak, AI often mixes lipsync or timing.
 **Solution:** ONE active speaker per shot.
 
-- **Shot A:** Character X speaks. Camera focuses on X (or over-the-shoulder of Y).
-- **Shot B:** Character Y replies. Camera focuses on Y.
+- **Shot A:** Character X speaks. Camera focuses on X (or over-the-shoulder of Y).
+- **Shot B:** Character Y replies. Camera focuses on Y.
+- Keep the same `speakerKey` and `voiceCast` identity in both shots. Only `performanceNote` changes.
 
 **EXCEPTION:** If both MUST be in frame (Two-Shot):
 1. Use "Reaction Shot" for the listener (listener nods while speaker talks off-screen sound).

package/content/skills/prompt-structure/SKILL.md

@@ -81,6 +81,18 @@ Before using any video template in this file:
 - If active model is **Kling 3.0**, do **not** default to the Veo-style `Audio direction:` bullet schema below.
 - If project is a hybrid/smart-hybrid package, the local `.agent/skills/prompt-structure/SKILL.md` override takes precedence over this base file.
 
+## Kling Storyboard / Multi-Shot Decision Rule
+
+The official Kling app surface exposes start/end control, optional end-frame anchoring, and custom storyboard-style shot prompting.
+Use that capability with restraint:
+
+- default to a single Start+End transition prompt when the beat is one continuous motion arc, one reveal, or one emotional turn
+- use custom storyboard only when the clip truly contains **2-3 distinct editorial phases** with different visual jobs
+- in this toolkit, cap app.kling custom storyboard mode at **3 custom shots** per video
+- never split one micro gesture, blink, glance, prop touch, or tiny head turn into separate storyboard shots
+- if a beat needs 4+ distinct phases, split into chained videos instead of bloating one generation
+- every storyboard shot must earn a clear job: establish / action / reveal / reaction / resolve
+
 ## Video Prompt Structure (Veo-Oriented Default Template)
 
 > Use this section when the active video model/routed section is **Veo 3.1**.  
@@ -198,9 +210,16 @@ Avoid: distorted faces, morphing, bad anatomy, extra limbs/fingers, blurry, flic
 | **Establishing** | 8s | 10–15s | Set scene, allow environment |
 | **Complex transformation** | N/A | 15s (max) | Multi-step with single-scene continuity |
 
-> **Kling Note:** Duration maps directly to **transformation budget**. More change between start and end frames = longer duration needed. See `.agent/model-profile.md` for details.
-
----
+> **Kling Note:** Duration maps directly to **transformation budget**. More change between start and end frames = longer duration needed. See `.agent/model-profile.md` for details.
+
+### Kling Anti-Fragmentation Rule
+
+- do not turn every sentence or gesture into a new Kling storyboard phase
+- if a single camera move can carry the beat cleanly, keep one shot and strengthen the motion path
+- if the beat reads as setup -> action -> settle, `first -> then -> finally` is usually enough
+- reserve multi-shot/storyboard mode for meaningful internal progression, not decorative complexity
+
+---
 
 ## Prompt Re-Take Strategy
 

package/content/workflows/chain.md

@@ -34,6 +34,7 @@ All generated files must keep `SHOTNN.md` naming.
 1. Read `$OUTPUT_DIR/shots/SHOT[last].md`.
 2. Extract `SON FRAME` description.
 3. Capture character/location continuity details.
+4. Reuse `$OUTPUT_DIR/shot-plan.json -> voiceCast` and keep every existing `speakerKey` stable.
 
 ### 3. Continue Generation
 
@@ -49,6 +50,7 @@ If model is kling-3.0: keep Start+End transition mode and first/then/finally mot
 - Write each shot to `$OUTPUT_DIR/shots/SHOT[NN].md`
 - Keep one-file-per-shot contract
 - Ensure `ILK/İLK FRAME` code block exists even when chained
+- Keep `Audio Plan` blocks aligned to the existing `voiceCast`
 - Update `$OUTPUT_DIR/_index.md`
 
 ### 5. Refresh Reports

package/content/workflows/finish.md

@@ -32,6 +32,8 @@ All shot files follow `SHOTNN.md` naming.
 5. Verify chain continuity.
 6. Verify Avoid line on every prompt.
 7. Verify `Model Control` block exists in every shot file.
+8. Verify `shot-plan.json` has `voiceCast` coverage for every speaker or narrator.
+9. Verify every speaking VIDEO section has an `Audio Plan` block with valid `activeSpeakerKey`.
 8. For `kling-3.0`, verify:
    - `Transition Mode: Start+End`
    - CFG value is documented
@@ -83,6 +85,7 @@ Do not declare completion unless:
 - delivery report is pass
 - final summary says pass
 - model-specific checks pass (Kling or Veo profile)
+- voice design contract passes (`voiceCast` + `Audio Plan` + single active speaker)
 
 ---
 

package/content/workflows/generate.md

@@ -30,7 +30,7 @@ Generate production-ready shot prompts from scenario and **SAVE TO FILES**.
 ```text
 $OUTPUT_DIR/
 ├── project-info.md          # Scenario, characters, settings
-├── shot-plan.json           # Single-agent plan + policy contract
+├── shot-plan.json           # Single-agent plan + policy + voiceCast contract
 ├── shots/
 │   ├── SHOT01.md            # First shot (main + coverage)
 │   ├── SHOT02.md            # Second shot
@@ -62,7 +62,10 @@ $OUTPUT_DIR/
    - coverage strategy
    - model contract (`model`, `preset`, `transition_mode`)
    - `dialogue_name_policy` (`preserve-original-dialogue` or `anonymize-dialogue`)
-10. Ensure directories exist: `$OUTPUT_DIR/`, `$OUTPUT_DIR/shots/`, `$OUTPUT_DIR/reports/`.
+   - top-level `voiceCast`
+   - voice defaults (`single_active_speaker`, `music_default`, `subtitles_default`)
+10. Ensure every speaker or narrator has a stable `speakerKey` before shot writing.
+11. Ensure directories exist: `$OUTPUT_DIR/`, `$OUTPUT_DIR/shots/`, `$OUTPUT_DIR/reports/`.
 
 ### 2. Batch Strategy
 
@@ -83,9 +86,16 @@ For EACH shot:
 1. Analyze scene type (Dialogue / Action / Emotional / Establishing).
 2. Build dramaturgy behavior (objective, obstacle, stakes, subtext, beat turns).
 3. If model is `kling-3.0`, perform Kling Frame Prep (see below).
-4. Generate main shot prompts (`ILK/İLK FRAME`, `SON FRAME`, `VIDEO`).
+4. If model is `kling-3.0`, choose Kling execution mode:
+   - `single-transition` for one strong motion arc or reveal
+   - `custom-storyboard` only for 2-3 distinct editorial phases
+4. Reuse or create the correct `voiceCast` entry for any speaking character or narrator.
+5. Generate main shot prompts (`ILK/İLK FRAME`, `SON FRAME`, `VIDEO`).
 5. `ILK/İLK FRAME` section MUST always include a fenced code block.
 6. If chained, first-frame code block must explicitly state: `Use SHOT[prev]_END as exact first frame`.
+7. Write a machine-readable `Audio Plan` JSON block for every VIDEO section.
+8. If dialogue or voiceover exists, require `activeSpeakerKey`, `dialogueLines`, and `performanceNote`.
+9. Keep one active speaker per shot. Split reply dialogue across multiple shots.
 7. Enforce hard quality floor:
    - `ILK FRAME`: minimum 80 words
    - `SON FRAME`: minimum 80 words
@@ -98,11 +108,11 @@ For EACH shot:
    - explicit eyeline target / body orientation when a subject looks at someone or something
    - explicit shared light source / bounce logic when multiple subjects share frame
    - explicit depth/scale integration when more than one plane is visible
-9. Generate coverage prompts (2-3 per main shot, min 70 words each).
-10. Add Turkish summary for shot and each coverage section.
-11. Apply model-specific generation gates (see below).
-12. Write shot file: `$OUTPUT_DIR/shots/SHOT[NN].md`.
-13. Update `$OUTPUT_DIR/_index.md`.
+10. Generate coverage prompts (2-3 per main shot, min 70 words each).
+11. Add Turkish summary for shot and each coverage section.
+12. Apply model-specific generation gates (see below).
+13. Write shot file: `$OUTPUT_DIR/shots/SHOT[NN].md`.
+14. Update `$OUTPUT_DIR/_index.md`.
 
 #### Kling Frame Prep (when model is kling-3.0)
 
@@ -113,9 +123,11 @@ Before writing prompts, design the Start→End transition:
     - 5s = 1 major change
     - 10s = 2-3 staged changes
     - 15s = complex multi-step
-3.  **Motion timeline:** Write 2-4 steps: `first → then → finally`.
-4.  **Face/hands stability:** Match orientations between start and end — avoid >45° face rotation.
-5.  **Camera safety:** Use only safe movements (slow push-in, pan, tilt, micro-sway, tripod-locked).
+3.  **Execution mode:** Default to `single-transition`; use `custom-storyboard` only when the shot truly has 2-3 meaningful internal phases.
+4.  **Motion timeline:** Write 2-4 steps: `first → then → finally`.
+5.  **Face/hands stability:** Match orientations between start and end — avoid >45° face rotation.
+6.  **Camera safety:** Use only safe movements (slow push-in, pan, tilt, micro-sway, tripod-locked).
+7.  **Anti-fragmentation:** Do not turn one glance, gesture, or prop touch into separate micro-shots. If custom storyboard is used, cap it at 3 stages and make each stage editorially distinct.
 
 #### Veo Gate (when model is veo31)
 
@@ -134,6 +146,8 @@ Before writing prompts, design the Start→End transition:
 - [ ] Start/end frames verified for same visual universe
 - [ ] Dialogue uses `"quotation marks"` with tone markers (Kling inline format)
 - [ ] Eyelines, plane map, and shared-light logic stay consistent across start/end frames
+- [ ] Defaulted to `single-transition` unless 2-3 distinct internal phases genuinely require custom storyboard
+- [ ] If custom storyboard is used: maximum 3 stages, no decorative micro-beats, and each stage changes function/framing/action
 
 ### 4. Validation Pass (Mandatory Before Completion)
 
@@ -190,6 +204,36 @@ Avoid: blurry, low-res, text, watermark, bad anatomy, distorted face ...
 Avoid: blurry, low-res, text, watermark, bad anatomy, distorted face ...
 ```
 
+### AUDIO PLAN
+
+```json
+{
+  "language": "...",
+  "type": "...",
+  "activeSpeaker": "...",
+  "activeSpeakerKey": "...",
+  "dialogueLines": [],
+  "dialogueTranscript": "...",
+  "performanceNote": "...",
+  "delivery": {
+    "pace": "...",
+    "intensity": "...",
+    "emotion": "...",
+    "projection": "...",
+    "stability": "..."
+  },
+  "sfx": [],
+  "ambience": [],
+  "music": null,
+  "mixTarget": {
+    "dialogue": 0,
+    "sfx": 0,
+    "ambience": 0
+  },
+  "hasSubtitles": false
+}
+```
+
 ### VIDEO
 
 #### Veo Format:
@@ -240,6 +284,10 @@ Avoid: ...
 Avoid: ...
 ```
 
+```json
+[Coverage audioPlan JSON block - voice binding only when coverage actually contains speech]
+```
+
 ### SHOT[NN]B - [Type]
 [Same format]
 

package/content/workflows/recover.md

@@ -14,15 +14,17 @@ $ARGUMENTS
 - delivery report is fail
 - missing required shot sections
 - continuity mismatch between neighboring shots
+- missing `voiceCast` entry or broken `activeSpeakerKey` binding
 
 ## Recovery Steps
 
 1. Identify failed files and sections.
 2. Fix only affected shots first.
 3. Regenerate neighboring shots only if continuity requires it.
-4. Re-run `/safety-check`.
-5. Regenerate `$OUTPUT_DIR/reports/DELIVERY-REPORT.md`.
-6. Update `$OUTPUT_DIR/_index.md` with recovered status.
+4. Repair `voiceCast` or `Audio Plan` bindings before rerunning reports.
+5. Re-run `/safety-check`.
+6. Regenerate `$OUTPUT_DIR/reports/DELIVERY-REPORT.md`.
+7. Update `$OUTPUT_DIR/_index.md` with recovered status.
 
 ## Exit Criteria
 

package/content/workflows/safety-check.md

@@ -38,6 +38,10 @@ Validate all prompts before delivery to ensure platform compliance.
 - coverage has 2-3 sub-shots in same file
 - all prompts include Avoid line
 - all video prompts include full audio direction block
+- `shot-plan.json` contains `voiceCast` when any shot has dialogue or narration
+- every speaking VIDEO section includes an `Audio Plan` block
+- every `activeSpeakerKey` resolves to `shot-plan.json -> voiceCast`
+- no dialogue shot contains more than one active speaker
 
 3. Continuity checks
 - `SHOT[N]_END` aligns with `SHOT[N+1]_START`

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@milenyumai/film-kit",
-  "version": "1.3.0",
+  "version": "1.4.1",
   "description": "Hollywood-standard cinematic prompt engineering toolkit with model profiles (Veo 3.1 / Kling 3.0). Auto-configures AI agents (Cursor, Claude Code, VS Code Copilot, Antigravity) with production-grade shot generation system.",
   "type": "module",
   "main": "./build/index.js",