Play 43 — AI Video Generation

AI-powered video generation and editing — text-to-video, image-to-video, async batch processing with priority queues, C2PA watermarking, prompt and frame content safety, and cost-controlled preview-first workflow.

Architecture

Component	Azure Service	Purpose
Generation Model	Azure OpenAI (video endpoint)	Text-to-video and image-to-video generation
Prompt Enhancement	Azure OpenAI (GPT-4o-mini)	Enhance prompts with cinematic details
Content Safety	Azure Content Safety	Prompt moderation + frame-by-frame video check
Job Queue	Azure Service Bus	Batch processing with priority queues
Video Storage	Azure Blob Storage (Cool tier)	Generated video output with 30-day retention
Orchestrator	Azure Container Apps	Async generation, polling, watermarking
Secrets	Azure Key Vault	API keys, connection strings

📐 Full architecture details

Aspect	Play 36 (Multimodal Agent)	Play 43 (Video Generation)	Play 49 (Creative AI)
Input	Images + text for analysis	Text/image prompts for generation	Multi-modal content brief
Output	Structured analysis	Generated video files (.mp4)	Creative assets (images, copy)
Processing	Real-time inference	Async (30-120s per video)	Near real-time
Safety	Image content safety	Frame-by-frame video safety + C2PA	Brand safety + content policy
Queuing	N/A	Service Bus with priority	N/A
Cost	~$0.01/request	$0.06-1.50/video	~$0.05/asset

DevKit Structure


43-ai-video-generation/
├── agent.md                              # Root orchestrator with handoffs
├── .github/
│   ├── copilot-instructions.md           # Domain knowledge (<150 lines)
│   ├── agents/
│   │   ├── builder.agent.md              # Pipeline + batch queue + watermark
│   │   ├── reviewer.agent.md             # Content safety + copyright + C2PA
│   │   └── tuner.agent.md                # Quality/cost + resolution + queue
│   ├── prompts/
│   │   ├── deploy.prompt.md              # Deploy pipeline + queue + safety
│   │   ├── test.prompt.md                # Generate test videos
│   │   ├── review.prompt.md              # Audit safety + copyright
│   │   └── evaluate.prompt.md            # Measure quality + cost
│   ├── skills/
│   │   ├── deploy-ai-video-generation/   # Full deploy with Service Bus + safety
│   │   ├── evaluate-ai-video-generation/ # Quality, safety, performance, cost
│   │   └── tune-ai-video-generation/     # Presets, prompts, queue, cost tuning
│   └── instructions/
│       └── ai-video-generation-patterns.instructions.md
├── config/                               # TuneKit
│   ├── openai.json                       # Resolution presets, prompt enhancement
│   ├── guardrails.json                   # Content safety, cost controls, C2PA
│   └── agents.json                       # Queue config, rate limits, storage
├── infra/                                # Bicep IaC
│   ├── main.bicep
│   └── parameters.json
└── spec/                                 # SpecKit
    └── fai-manifest.json

Quick Start


# 1. Deploy video generation pipeline
/deploy
 
# 2. Generate test videos
/test
 
# 3. Audit content safety and copyright
/review
 
# 4. Measure video quality and cost
/evaluate

Key Metrics

Metric	Target	Description
Prompt Adherence	> 4.0/5.0	Video matches prompt description
Content Safety	> 99%	Frames pass Content Safety check
C2PA Watermark	100%	AI-generated metadata embedded
Generation Latency (5s)	< 60s	Time from request to video ready
Cost per 5s/1080p	< $0.50	API + safety + storage
Queue Throughput	> 20/hr	Videos generated per hour

Estimated Cost

Service	Dev/mo	Prod/mo	Enterprise/mo
Azure OpenAI	$50	$350	$1,200
Azure Container Apps (GPU)	$60	$500	$1,800
Blob Storage	$5	$60	$200
Azure AI Content Safety	$0	$80	$250
Service Bus	$0	$10	$80
Azure CDN	$5	$40	$150
Key Vault	$1	$5	$15
Application Insights	$0	$25	$80
Total	$121	$1,070	$3,775

Estimates based on Azure retail pricing. Actual costs vary by region, usage, and enterprise agreements.

💰 Full cost breakdown

WAF Alignment

Pillar	Implementation
Responsible AI	C2PA watermarking, frame-by-frame safety, copyright filtering
Security	Content Safety on prompts and output, Key Vault for secrets
Cost Optimization	Preview-first workflow, resolution presets, Cool storage tier
Reliability	Service Bus with retry, dead letter queue, 3-attempt retry
Performance Efficiency	Async generation, priority queues, concurrent processing
Operational Excellence	Job tracking, cost estimation, rate limiting, daily budget caps

FAI Manifest

Field	Value
Play	`43-ai-video-generation`
Version	`1.0.0`
Knowledge	F1-GenAI-Foundations, F2-LLM-Selection, T2-Responsible-AI, T3-Production-Patterns, R1-Prompt-Patterns
WAF Pillars	cost-optimization, responsible-ai, performance-efficiency, reliability
Groundedness	≥ 85%
Safety	0 violations max