Logo

Getting Started

  • Installation & Setup
    • Prerequisites
    • Clone the Repository
    • Environment Configuration
    • Quick Setup (Recommended)
    • Manual Setup
      • 1. Start Docker Containers
      • 2. Install Dependencies
      • 3. Run Migrations
      • 4. Create a Superuser
      • 5. Seed the Database
      • 6. Start the Application
    • Process Architecture
    • Verifying the Setup
    • Optional: Pre-download Models
    • Next Steps
  • Your First Image
    • Overview
    • Step 1: Create a Prompt
    • Step 2: Create a DiffusionJob
    • Step 3: Watch It Generate
    • Step 4: View the Result
    • Experimenting Further
    • Next Steps
  • Your First Adaptation
    • Overview
    • Step 1: Create a Brand
    • Step 2: Create a Campaign
    • Step 3: Create the Origin VideoAdUnit
      • Adding Script Rows
    • Step 4: Create an Adaptation
      • The Pipeline Runs Automatically
      • Viewing the Results
    • Step 5: Generate a Storyboard
    • What Just Happened
    • Next Steps

Concepts

  • Marketer’s Guide to Gen AI
    • The Talent: The Model (or “Checkpoint”)
    • The Account Exec: CLIP
    • The Process: UNet
    • The Delivery: VAE
    • The Specialists (The Add-ons)
      • The Brand Cop: LoRA
      • The Art Director: ControlNet
      • The Mood Board: IP-Adapter
    • The Production Studio: Video & Audio
      • The Continuity Director: Temporal Consistency (Video AI)
      • The Sound Engineer: Generative Audio
    • Equity & Legal Rules of Engagement
      • The Creative Director is Human (generative AI has no sense of taste)
      • Rights Management: Respect & Compensation Still Matter
    • The Frontier: Building Your Own Lab
      • The Traffic Dept: TVC Adaptation
  • Architecture Overview
    • Process Model
      • Docker Services
    • Project Structure
      • Django Apps
      • Library Modules
    • Data Flow
      • Image Generation
      • TV Spot Adaptation
    • Configuration Layers
    • Celery Task Architecture
  • Diffusion Models
    • Design Goals
    • Template Method Pattern
      • load_pipeline()
      • generate()
      • Hook Methods
    • Mixins
      • CompelPromptMixin
      • CLIPTokenLimitMixin (Legacy)
      • DebugLoggingMixin
    • Concrete Models
    • ModelFactory
    • Configuration Flags
    • Warm Cache
    • Device Optimization
    • Model Comparison
    • Adding a New Model
  • Adaptation Pipeline
    • Overview
    • Graph Structure
    • Pipeline Nodes
      • Concept Analyst
      • Cultural Researcher
      • Writer
      • Format Evaluator
      • Cultural Evaluator
      • Concept Evaluator
      • Brand Evaluator
    • Evaluation Gates
    • Model Resolution
    • PipelineModelLoader
    • Output Artifacts
  • Audience Targeting
    • Data Model
    • Geographic Hierarchy
    • Non-Geographic Segments
    • Personas
    • Insights
    • Insights Composition
    • How It Feeds the Pipeline
  • Observability
    • Logging Architecture
      • Log Files
      • JSON Format
      • Loggers
    • Grafana + Loki
      • Querying Logs
      • Local Log Analysis
    • Flower Task Monitor
    • Debugging a Failed Job

Guides

  • Adding a New Model
    • Prerequisites
    • Step 1: Create the Model File
    • Step 2: Register in the Factory
    • Step 3: Add to presets.json
    • Step 4: Sync to Database
    • Adding Behavior with Hooks
    • Adding Mixins
    • Configuration Flags
  • Working with LoRAs
    • Overview
    • Adding a LoRA via presets.json
      • Configuration Fields
    • Architecture Compatibility
    • CivitAI Auto-Download
      • AIR URN Format
      • Setup
      • Download Flow
    • Trigger Words
    • Strength and Overrides
    • Local File Resolution
    • Using LoRAs in the Admin
  • Prompt Enhancement
    • Overview
    • Rule-Based Enhancement
    • HuggingFace Local Model
    • Anthropic API
    • Style Options
    • Creativity Level
    • Integration with DiffusionJob
      • VRAM Management
    • Configuring Enhancement
  • Compel Prompt Weighting
    • What Is Compel?
    • Supported Models
    • Prompt Weighting Syntax
      • Examples
    • Long Prompt Support
    • LoRA Integration
    • Practical Tips
  • Creating a Campaign
    • What Is a Campaign?
    • Creating a Campaign in the Admin
    • Linking a Brand
    • Setting Original Script Data
    • Campaign Lifecycle
    • Viewing Campaign Status
  • Origin & Adaptation Ad Units
    • Ad Unit Model
    • Creating an Origin Ad Unit
    • Creating an Adaptation Ad Unit
    • Pipeline Status Lifecycle
    • Per-Node Model Configuration
    • Brand Override
    • Pipeline Output
    • Visual Style Prompt
  • Importing TV Spots (JSON)
    • JSON Format
      • Required Fields
      • Optional Fields
      • Script Row Fields
      • Example JSON
    • The import_tvspot Command
      • Usage
      • What the Command Does
      • Prerequisites
      • Validation Rules
      • Dry Run Output
      • Import Output
    • Error Handling
    • After Import
  • Video Upload & Analysis
    • Overview
    • Uploading a Video
    • Security Validation
    • Analysis Phases
      • Phase 1: Extract Metadata
      • Phase 2: Scene Detection
      • Phase 3: Audio Transcription
      • Phase 4: Keyframe Extraction & Object Detection
      • Phase 5: Visual Style Analysis
      • Phase 6: Camera Work Analysis
      • Phase 7: Sentiment Analysis
      • Phase 8: Scene Categorization
      • Phase 9: Script Generation
      • Phase 10: Audience Insights
    • VideoProcessingResult
      • KeyFrame Records
    • AdUnitMedia Lifecycle
    • Viewing Results
  • Storyboard Generation
    • Overview
    • Creating a Storyboard
    • Prompt Generation
      • Visual Element Extraction
      • LLM Enhancement
    • DiffusionJob Creation
    • Task Flow
    • Progress Tracking
      • Status Lifecycle
    • VRAM Management
    • Viewing Storyboard Frames
      • Comparing Configurations
  • Regions, Countries & Languages
    • Data Model
      • Region
      • Country
      • Language
      • Many-to-Many Relationships
    • Adding a New Market
    • Language-to-LLM Model Linking
    • Insights Structure
    • Import & Export
  • Segments & Personas
    • Segments
      • Creating a Segment
      • Segment Insights
    • Personas
      • Creating a Persona
      • Cascading Admin Selectors
    • Linking Personas to Ad Units
    • How Personas Feed the Pipeline
    • Import & Export
  • Brand Configuration
    • Brand Model
      • Creating a Brand
    • How Brand Evaluation Works
    • Linking Brands
      • Campaign Level
      • Ad Unit Level (Override)
    • Guidelines vs. Insights
    • Import & Export
  • Prompt Template Editing
    • Overview
    • Editing a Template
    • Template Categories
    • Jinja2 Syntax
    • Template Variables
    • Versioning
    • Cache Invalidation
    • Usage Analytics
    • Import & Export
  • Per-Node Model Selection
    • Model Resolution Chain
    • Node Keys
    • Setting Overrides at the Ad Unit Level
    • Setting Defaults via PipelineSettings
    • Practical Examples
    • How Resolution Works
  • Pipeline Settings
    • Overview
    • Accessing PipelineSettings
    • Configurable Fields
    • How Settings Interact with Overrides
    • Configuration Strategies
      • Uniform Model
      • Split by Role
      • Quality vs. Speed
    • Editing in the Admin
    • LLM Models
  • Data Import & Export
    • Command Overview
      • Diffusion
      • Core
      • Audiences
      • TV Spots
    • Import Dependency Order
    • Common Options
      • --dry-run
      • --dir
    • Data Files
    • Backup & Restore
  • Grafana & Log Querying
    • Accessing Grafana
    • LogQL Query Examples
      • Basic Queries
      • Filtering by Level
      • Text Search
      • Filtering by Logger
      • Correlating Task IDs
    • Flower Task Monitor
    • Local Log Analysis
      • Streaming Logs
      • Filtering by Level
      • Filtering by Logger
      • Text Search
    • Log Files
  • Troubleshooting
    • Setup & Environment
      • ModuleNotFoundError: 'cw'
      • DJANGO_SETTINGS_MODULE not set
      • Missing static files in admin
      • Migration errors
    • Model Loading
      • Out of memory (OOM)
      • Model not found
      • Wrong dtype
    • LoRA Issues
      • Architecture mismatch
      • Missing trigger words
      • CivitAI download failures
    • Celery & Task Issues
      • Tasks stuck in “pending”
      • Worker crashes
      • Task timeouts
    • Adaptation Pipeline
      • Evaluation gate loops
      • Schema validation errors
      • Pipeline model loading errors
    • Docker & Infrastructure
      • Port conflicts
      • Container startup order
    • Diagnostic Commands

Reference

  • Management Commands
    • Diffusion
    • Core
    • Audiences
    • TV Spots
    • Common Flags
      • --dry-run
      • --dir
      • --file
    • Import Dependency Order
    • Idempotency
  • Data Schemas (JSON)
    • presets.json
    • llm_models.json
    • regions.json
    • countries.json
    • languages.json
    • country_regions.json
    • country_languages.json
    • prompt_templates.json
    • segments.json
    • personas.json
    • persona_segments.json
    • brands.json
    • example_tvspot.json
    • Insights Schema
    • Relationship Diagram
  • Model Specifications
    • Summary Table
    • Z-Image Turbo
    • Flux.1 Dev
    • Qwen-Image-2512
    • SDXL Turbo
    • Juggernaut XL v9
    • DreamShaper XL Lightning
    • Realistic Vision v5.1
    • Architecture Compatibility
    • Token Handling
    • Device Optimizations
  • Configuration
    • Environment Variables
      • Database
      • Broker
      • Application
      • Video Upload
    • Presets Configuration
      • Top-Level Structure
      • Model Fields
      • Model Settings
      • Behavior Flags
      • LoRA Fields
      • LoRA Settings
    • Django Settings
      • Celery
      • Logging
      • Video Upload
    • Docker Compose Services
  • Celery Tasks & Queues
    • Task Reference
      • enhance_prompt_task
      • generate_images_task
      • download_lora_task
      • create_adaptation_task
      • generate_storyboard_task
      • analyze_video_task
    • Queue Configuration
    • Worker Configuration
      • Pool
      • Time Limits
      • Result Storage
      • Broker
    • Model Caching
    • Retry Behavior
    • Starting Workers
  • API Reference
    • Core
      • Models
        • LLMModel
        • PromptTemplate
        • PipelineSettings
      • Admin
        • LLMModelAdmin
        • PromptTemplateAdmin
        • PipelineSettingsAdmin
      • Prompt Rendering
        • render_prompt()
    • Diffusion
      • Models
        • DiffusionModel
        • LoraModel
        • ControlNetModel
        • Prompt
        • DiffusionJob
      • Tasks
      • Admin
        • UserAdmin
        • GroupAdmin
        • TaskResultAdmin
        • GroupResultAdmin
        • JobInline
        • DiffusionModelAdmin
        • ControlNetModelAdmin
        • LoraModelAdmin
        • PromptStatusFilter
        • PromptAdmin
        • DiffusionJobAdmin
    • TV Spots
      • Models
        • Brand
        • Campaign
        • AdUnit
        • VideoAdUnit
        • AdUnitScriptRow
        • Storyboard
        • StoryboardImage
        • AdUnitMedia
        • VideoProcessingResult
        • KeyFrame
      • Tasks
      • Admin
        • BrandAdmin
        • AdUnitMediaInline
        • VideoAdUnitInline
        • CampaignAdmin
        • AdUnitScriptRowInline
        • VideoAdUnitAdmin
        • StoryboardImageInline
        • StoryboardAdmin
        • KeyFrameInline
        • VideoProcessingResultAdmin
        • AdUnitMediaAdmin
    • Audiences
      • Models
        • Region
        • Country
        • Language
        • CountryRegion
        • CountryLanguage
        • LanguageAlternativeModel
        • Segment
        • Persona
        • PersonaSegment
        • WVSProfile
      • Admin
        • InsightsWidgetMixin
        • LanguageAlternativeModelInline
        • CountryRegionInline
        • CountryLanguageInline
        • PersonaSegmentInline
        • PersonaSegmentsSection
        • LanguageAdmin
        • RegionAdmin
        • CountryAdmin
        • SegmentAdmin
        • PersonaAdmin
        • WVSProfileAdmin
      • Views
        • get_segment_vectors()
        • get_segment_values()
        • add_segment_to_persona()
        • remove_segment_from_persona()
    • Library Modules
      • Diffusion Models
        • Model Factory
        • Base Model
        • Mixins
      • Adaptation Pipeline
        • State & Initialization
        • Pipeline Nodes
        • Graph Construction
        • Model Loader
      • Prompt Enhancement
        • PromptEnhancer
        • HFPromptEnhancer
        • LLMPromptEnhancer
        • process_prompts_from_file()
        • main()
      • LoRA Management
        • LoRAManager
      • Configuration
        • PresetsConfig
        • get_config()
      • CivitAI Integration
        • parse_air()
        • fetch_model_version_metadata()
        • extract_lora_metadata()
        • download_lora()
      • Insights Composition
        • compose_insights()
        • compose_insights_as_markdown()
      • Storyboard Generation
        • StoryboardGenerator
        • create_storyboard_jobs()
        • WireframePromptBuilder
        • create_wireframe_storyboard_jobs()
      • Video Analysis
        • extract_video_metadata()
        • detect_scenes()
        • transcribe_audio()
        • detect_objects()
        • detect_objects_batch()
        • summarize_objects()
        • analyze_camera_work()
        • analyze_lighting()
        • analyze_visual_style()
        • extract_dominant_colors()
        • analyze_sentiment()
        • categorize_scenes()
        • summarize_categories()
        • generate_audience_insights()
      • File Security
        • FileSecurityValidator
        • FileSizeValidator
        • FileExtensionValidator
        • MimeTypeValidator
        • FileHeaderValidator
        • FilenameSanitizer
        • VideoFileValidator

Project

  • Philosophy & Design Principles
    • Vision
    • Principles
      • Creative AI Augments, Not Replaces
      • Multi-Model by Design
      • Cultural Sensitivity as a First-Class Concern
      • Configuration Over Code
      • Template Method for Extensibility
      • Sequential Execution for GPU Safety
      • Database as Source of Truth
    • Technical Foundations
  • Contributing
    • Contributing to Generative Creative Lab
      • Welcome
      • Development Setup
        • Prerequisites
        • Initial Setup
      • Code Style
        • Python Style
        • Import Ordering
        • Docstring Format
        • Type Hints
      • Making Changes
        • Fork and Branch Workflow
        • Branch Naming Conventions
        • Commit Message Format
        • Testing Requirements
      • Pull Request Process
        • Before Submitting
        • PR Title and Description
        • Review Process
        • Merge Requirements
      • Issue Guidelines
        • Bug Reports
        • Feature Requests
        • Documentation Issues
      • Code of Conduct
        • Our Standards
        • Unacceptable Behavior
        • Enforcement
      • Getting Help
        • Where to Ask Questions
        • Response Time Expectations
      • Additional Resources
      • License
  • Roadmap
    • Completed
      • Core Platform
      • TV Spot Adaptation
      • Reference Data
      • Infrastructure
    • Current Priorities
    • Future Ideas
    • Non-Goals
  • Research Notes
    • Multilingual LLM Model Selection
      • Evaluated Models
      • Selection Criteria
    • Adaptation Pipeline Design
      • Why Separate Evaluation Gates
      • Retry Budget
      • Structured Generation
    • Video Upload Security
      • Threat Model
      • Validation Chain
    • Model Memory Management
      • Sequential Execution
      • Eviction Strategy
      • Quantization
  • Changelog
    • Changelog
      • [Unreleased]
        • Added
        • Changed
        • Fixed
        • Removed
        • Migration Notes
      • Version History
  • Non-Commercial License
    • 1. Definitions
    • 2. License Grant
    • 3. Non-Commercial Use Only
    • 4. Distribution Requirements
    • 5. Restrictions
    • 6. Intellectual Property
    • 7. Outputs and Model Usage
    • 8. Term and Termination
    • 9. Disclaimers
    • 10. Limitation of Liability
    • 11. Governing Law and Venue
    • 12. Changes to This License
Generative Creative Lab
  • Search

© Copyright 2026, Andrew Marconi.

Built with Sphinx using a theme provided by Read the Docs.