Overview

Playbook 03 Steering & Authoring Guide: Autonomous Agentic Video Studio Company (Kids Karaoke & Educational Songs)

Playbook ID: PB-03
Topic: Autonomous Multi-Agent AI Studio Architecture — Operating a Virtual Production Company to Produce Children's Educational Karaoke & Sing-Along English Learning Videos
Tooling Stack: Gemini 2.5 Flash & Pro (Reasoning & Orchestration), Google Veo 2 / Veo 3 (veo-2.0), Imagen 3 (imagen-3.0-generate-002), Google Cloud TTS Journey Voices (en-US-Journey-F), DeepMind Lyria / MusicFX, FFmpeg 7.0+
Target Audience: Year 1 Computer Science & Software Engineering Students
Delivery Standard: Zero Fluff, 7 Universal Quality Acceptance Gates, Mandatory Hands-On Labs, Executable Tested Solutions
Publication Target: Markdown (.md) pushed to GitHub (origin/main) for local review on D:\Agent-assisted-research. (Word/Google Docs export compiled only upon explicit user approval).


1. Executive Purpose & Business Mission

The objective of Playbook 03 is to provide a complete, production-grade architectural blueprint for designing, deploying, and running an Autonomous Multi-Agent AI Video Studio.

This virtual studio operates as an automated production company that ideates, scripts, composes, directs, animates, quality-audits, packages, and publishes educational karaoke videos for children (ages 2–6) to sing English songs and acquire foundational language skills (vocabulary, phonics, rhythm, and sentence structures).

Unlike single-prompt toys or disconnected scripts, this playbook establishes a resilient, stateful multi-agent system with specialized personas, blackboard state management, automated rejection loops, and deterministic post-production pipelines.

┌────────────────────────────────────────────────────────────────────────┐
│             VIRTUAL AI VIDEO STUDIO: AGENT ORGANIZATIONAL CHART        │
├────────────────────────────────────────────────────────────────────────┤
│                     EXECUTIVE PRODUCER AGENT                           │
│   (Curriculum Roadmap, Production Schedule, Token & Cost Budgeting)    │
│                                │                                       │
│      ┌─────────────────────────┴─────────────────────────┐             │
│      ▼                                                   ▼             │
│ ┌───────────────────────────┐         ┌──────────────────────────────┐ │
│ │  PEDAGOGICAL CURRICULUM   │         │   MUSIC & LYRICIST AGENT     │ │
│ │  - CEFR Pre-A1 phonics    │         │   - 108 BPM nursery songs    │ │
│ │  - Vocabulary scaffolding │         │   - Syllable meter mapping   │ │
│ └─────────────┬─────────────┘         └──────────────┬───────────────┘ │
│               │                                      │                 │
│               └──────────────────┬───────────────────┘                 │
│                                  │                                     │
│      ┌───────────────────────────┴───────────────────────────┐         │
│      ▼                                                       ▼         │
│ ┌───────────────────────────┐         ┌──────────────────────────────┐ │
│ │  VISUAL ASSET & MASCOT    │         │   ANIMATION DIRECTOR AGENT   │ │
│ │  - Imagen 3 turnarounds   │         │   - Veo 2 shot choreography  │ │
│ │  - Invariant tokens       │         │   - Cognitive stillness rule │ │
│ └─────────────┬─────────────┘         └──────────────┬───────────────┘ │
│               │                                      │                 │
│               └──────────────────┬───────────────────┘                 │
│                                  │                                     │
│      ┌───────────────────────────┴───────────────────────────┐         │
│      ▼                                                       ▼         │
│ ┌───────────────────────────┐         ┌──────────────────────────────┐ │
│ │  POST-PRODUCTION & KARAOKE│         │   QUALITY & COMPLIANCE GATE  │ │
│ │  - ASS bouncy subtitles   │◄───────►│   - COPPA / Child safety QA  │ │
│ │  - Dynamic audio ducking  │ (Reject)│   - Syllable sync audit      │ │
│ └─────────────┬─────────────┘         └──────────────────────────────┘ │
│               │                                                        │
│               ▼                                                        │
│ ┌────────────────────────────────────────────────────────────────────┐ │
│ │                PUBLISHING & DISTRIBUTION AGENT                     │ │
│ │   - YouTube Kids metadata, thumbnails, packaging, metrics tracking  │ │
│ └────────────────────────────────────────────────────────────────────┘ │
└────────────────────────────────────────────────────────────────────────┘

2. Mandatory Tool Currency & Horizon Review Protocol

In accordance with strict production guidelines, every agent architecture, prompt recipe, and runnable code lab in Playbook 03 must use the latest active Google and industry tools:

  1. Gemini 2.5 Flash & Gemini 2.5 Pro:
    • gemini-2.5-flash: Rapid agent-to-agent message passing, JSON schema enforcement, triage, and high-frequency linting.
    • gemini-2.5-pro: Complex pedagogical curriculum design, deep reasoning, musical meter analysis, and multi-modal visual QA.
    • Never reference legacy model names (gemini-1.5-flash, gemini-2.0-flash-exp).
  2. Google Veo 2 / Veo 3 (veo-2.0):
    • Production spatio-temporal DiT video generation.
    • Strictly Image-to-Video (I2V) first-frame conditioning.
    • Enforce 24fps constant frame rate and toddler camera safety rules (velocity <= 0.05m/s, zero camera roll).
  3. Google Imagen 3 (imagen-3.0-generate-002):
    • Stylized 3D Pixar / claymation cartoon aesthetics (anti-uncanny valley).
    • 4-panel orthographic turnaround model sheets for character consistency.
  4. Google Cloud TTS Journey Voices (en-US-Journey-F):
    • Infant-Directed Speech (IDS) prosody: 80%-85% rate, +12% to +15% fundamental frequency lift.
    • Programmatic <break time="2500ms"/> child verbal response windows.
  5. Google DeepMind Lyria / MusicFX:
    • 108 BPM nursery song generation in pure diatonic major keys (C Major, G Major).
    • Lightweight acoustic orchestration (marimba, ukulele, pizzicato strings, glockenspiel).
  6. FFmpeg 7.0+ Automated Post-Production Engine:
    • Advanced SubStation Alpha (.ass) karaoke bouncy syllable highlighting ({\k<duration>} tags).
    • Dynamic sidechain ducking (-12dB background music during vocals).

3. The 7 Universal Quality Acceptance Gates

Every chapter of Playbook 03 must satisfy all 7 acceptance gates before publication:

  1. Gate 1: Zero Fluff & Agentic Engineering Rigor: Open directly with agent architecture, state transitions, message schemas, and cognitive models. No generic motivational intros.
  2. Gate 2: Mandatory Naive vs. Production Contrasts: Directly contrast naive single-prompt monolithic pipelines with production multi-agent state machines featuring checkpointing and rejection gates.
  3. Gate 3: Latest Google Model Configurations & Schemas: Detail exact parameter dictionaries for Gemini 2.5, Veo 2, Imagen 3, Cloud TTS Journey, and Lyria.
  4. Gate 4: Quantitative Trade-Off Matrix: Benchmark agent topologies, token economics, latency, cost per finished song episode, and automated failure recovery rates.
  5. Gate 5: The 10 Operational Failure Modes in Agentic Media Studios: Address multi-agent gotchas (infinite loops, semantic drift across agent handoffs, quota exhaustion cascades, state corruption, subtitle phase desynchronization).
  6. Gate 6: Mandatory Hands-On Lab (Interactive Challenge): Every chapter must feature an actionable, step-by-step challenge allowing Year 1 students to run and test that agent's subsystem directly on their computer.
  7. Gate 7: Mandatory Recommended Answer & Executable Solution: Every lab must provide a fully tested, runnable zero-dependency Python 3.11+ script with built-in test assertions certifying 100% compliance.

4. Master 8-Chapter Syllabus

Chapter Title Agent Role & Architecture Focus Hands-On Lab (Gate 6) Recommended Solution (Gate 7)
Ch 01 Multi-Agent Architecture & Virtual Studio Operating Model Role specialization, blackboard state architecture, directed acyclic graph (DAG) execution, studio token economics. Virtual Studio Agent Topology & Blackboard Engine. Python 3.11+ zero-dep multi-agent event bus & state machine simulator.
Ch 02 The Executive Producer & Curriculum Agent: Song Selection & Phonics CEFR Pre-A1 nursery rhymes (Wheels on the Bus, Old MacDonald, ABC Song), syllabus scheduling, budget guardrails. Autonomous Song Syllabus & Token Budget Planner. Python curriculum optimizer mapping phonetic targets to 60s song structures.
Ch 03 The Music & Lyricist Agent: Rhythmic Syllable Stress & Lyria Songwriting 108 BPM nursery song generation, poetic meter, syllable duration calculation, vocal-to-instrumental stem generation. Rhythmic Syllable Stress & Lyria Prompt Compiler. Python meter analyzer aligning lyrics with musical beats & generating MusicFX prompts.
Ch 04 Visual Asset & Mascot Consistency Agent: Imagen 3 Fleet Multi-character turnaround sheets, style seed locking, prop generation, invariant token enforcement across shots. Multi-Character Mascot Fleet Consistency Engine. Parameterized Imagen 3 turnaround & asset manifest generator with consistency linting.
Ch 05 The Animation Director Agent: Google Veo 2 Scene Choreography Spatio-temporal DiT video directing, 3-second cognitive stillness, I2V keyframe conditioning, shot-by-shot kinematics. Song-Synchronized Video Choreography Engine. Python Veo 2 directing compiler converting musical verses into camera action payloads.
Ch 06 The Post-Production Audio & Bouncy Subtitle Agent FFmpeg 7.0+ sidechain ducking (-12dB), Advanced SubStation Alpha (.ass) karaoke syllable timing, video concatenation. Automated Karaoke Subtitle & Audio Ducking Pipeline. Python + FFmpeg orchestrator generating bouncy syllable .ass files and mixing audio.
Ch 07 The Quality Auditor & COPPA Compliance Gatekeeper Automated multi-modal verification, COPPA compliance, acoustic safety linting, rejection & repair loops. Autonomous Multi-Modal Studio Quality Gatekeeper. Python QA engine executing automated image, audio, and subtitle validation assertions.
Ch 08 Autonomous Studio Orchestrator & YouTube Kids Publishing Engine End-to-end orchestration: Song Idea -> Script -> Music -> Video -> Post -> YouTube packaging. Full CLI Virtual Kids Video Studio Company Suite. Runnable end-to-end Python 3.11+ orchestration script compiling a complete 60s karaoke video.
App A Studio Infrastructure, Agent Prompts & Production Economics Complete system prompt library for all 8 agents, SQLite database schemas, API cost calculators, quick-reference cheat sheets. Studio Infrastructure & Deployment Reference. Production sizing calculator, agent prompts, and operational runbooks.

5. Pedagogical Directing Rules for Kids Sing-Along Media

  1. Rhythmic Syllable Concordance: Every on-screen bouncing ball or glowing syllable must highlight on the exact acoustic beat of the singing voice (zero phase lag).
  2. High-Contrast Bouncy Subtitles: Use 64pt Arial Rounded MT Bold with canary yellow active highlight (&H0033FFFF&) and dark navy outline (&H004C2B1A&) placed in the lower third.
  3. The 108 BPM Golden Tempo: Keep song tempos locked at 100-112 BPM. Slower bores toddlers; faster prevents them from singing along.
  4. Mandatory Chorus Repetition: Essential vocabulary must repeat in identical musical phrases at least 3 times per 60-second song.
  5. Safe Cognitive Visual Pacing: Mascot movements during singing must follow gentle, predictable rhythms with held poses during key phonetic vowels.