Back to Portfolio

'Prompts, Robots, Action!': AI-Powered Movie Poster Generator

PythonOpenAI APIMachine LearningBERTData Analysis
'Prompts, Robots, Action!': AI-Powered Movie Poster Generator preview

An AI system that generates movie posters by combining GPT-4, DALL-E 3, and BERT for creative content generation and sentiment analysis.

Examples of the Generated Movie Posters

Click or tap a poster for its breakdown

Generated Movie Poster 1Generated Movie Poster 2Generated Movie Poster 3Generated Movie Poster 4

Status

Completed (May 2024)

Inspiration

This project emerged from exploring the intersection of AI and creative industries, specifically investigating how Large Language Models could contribute to visual content generation. With AI's growing influence in film production, I wanted to build a system that could demonstrate both technical sophistication and creative capability in generating movie marketing materials.

Technical Challenges & Solutions

API Integration & Rate Management

  • Implemented efficient handling of OpenAI API calls across multiple services (GPT-3.5/4, DALL-E 3)
  • Developed robust error handling and rate limiting system
  • Created fallback mechanisms for API failures to ensure system reliability

Data Pipeline Architecture

  • Designed modular system with four main components: Prompt Generator, Review Generator, Sentiment Analysis, and Poster Generator
  • Built logging system for tracking generated content using CSV for efficient data management
  • Implemented systematic content cataloging for analysis and comparison

Machine Learning Integration

  • Transitioned from TextBlob to BERT for more sophisticated sentiment analysis
  • Mapped emotional content to 100 distinct art styles using custom classification system
  • Integrated multiple ML models (GPT, BERT, DALL-E) into a cohesive workflow

Implementation Metrics

  • Generated and analyzed 50+ unique movie posters
  • Achieved 85% successful generation rate across all system components
  • Survey results (n=30) showed:
    • 3.4/5 average attention-grabbing score
    • 3.2/5 average creativity rating
    • 3.0/5 average marketing effectiveness
  • GPT-4 showed 15% improvement in generating engaging content compared to GPT-3.5

Skills Demonstrated

  • Programming: Python, API integration, error handling, data processing
  • ML/AI: OpenAI APIs (GPT-3.5, GPT-4, DALL-E 3), BERT implementation
  • Data: CSV handling, logging systems, metrics tracking
  • Research: Survey design, statistical analysis, user feedback collection
  • Project Management: Scope definition, iteration planning, documentation

Development Process

1. Research & Planning

  • Investigated existing AI art generation systems
  • Analyzed movie poster design principles
  • Defined success metrics and evaluation frameworks

2. Implementation

  • Built modular components with clear interfaces
  • Integrated multiple AI services
  • Created robust logging and tracking systems

3. Testing & Validation

  • Conducted user surveys using AIDA marketing model
  • Performed comparative analysis between GPT versions
  • Validated system outputs against design objectives

4. Documentation & Analysis

  • Created comprehensive technical documentation
  • Analyzed system performance and user feedback
  • Identified areas for future enhancement

Next Steps

  • Implement advanced prompt engineering for improved image generation
  • Expand art style training dataset for better aesthetic matching
  • Develop automated A/B testing system for poster effectiveness
  • Create API endpoint for external service integration
  • Add support for custom art style definitions

Key Features

  • Multi-model AI integration (GPT-3.5/4, DALL-E 3, BERT)
  • Robust error handling and rate limiting
  • Modular data pipeline architecture
  • Comprehensive logging and analytics system
  • Custom art style classification system