EF GenAI Hackathon - Winner
National Hackathon
Built a prompt-to-video engine with intelligent image selection, sentiment-based BGM, and multilingual support; winner from 200+ participants.
The Story
The EF GenAI Hackathon 2023 challenged participants to push the boundaries of generative AI. Our winning project was a comprehensive prompt-to-video engine that could transform text descriptions into engaging video content.
Core Features
Intelligent Image Selection:
The system analyzed prompts to understand visual requirements and automatically sourced or generated appropriate imagery. Using a combination of image databases and AI generation, it could match the tone and subject matter of any given prompt.
- *Sentiment-Based Background Music:*
- Perhaps the most innovative feature was our BGM system. By analyzing the emotional arc of the input text, the engine would:
- •Detect sentiment changes throughout the narrative
- •Select music tracks that matched the emotional tone
- •Dynamically adjust tempo and intensity to match scene transitions
- *Multilingual Support:*
- We built the engine to handle multiple languages from the ground up, enabling:
- •Text-to-speech in various languages
- •Culturally appropriate visual selections
- •Subtitle generation for accessibility
Technical Implementation
• Custom prompt engineering for consistent visual outputs • Audio analysis pipeline for music-mood matching • Modular architecture allowing easy language additions
Winning against 200+ participants validated our approach of focusing on the complete user experience rather than just technical novelty. The judges particularly appreciated how the different AI components worked together seamlessly.