Inspiration
Manga has a unique way of storytelling, combining visuals and emotions in a way that text alone cannot. We wanted to create a tool that bridges the gap between written stories and dynamic manga panels, making it easier for writers to visualize their narratives.
What it does
Mangify takes written text and converts it into manga-style panels. It extracts scene descriptions, character actions, and dialogues, then generates background, character expressions, and action details before assembling them into a cohesive manga panel.
How we built it
Text Processing: Extracts key elements like characters, setting, and dialogue.
Image Generation: Uses AI models to generate characters, backgrounds, and actions separately.
Composition Pipeline: Merges these elements while ensuring consistency across panels.
Speech Bubble Integration: Overlays text using an optimized placement algorithm.
Challenges we ran into
Maintaining character consistency across multiple panels.
Generating coherent action scenes from text prompts.
Ensuring speech bubbles are placed naturally in each panel.
Accomplishments that we're proud of
Successfully generating consistent character appearances across scenes.
Streamlining a multi-step AI pipeline into a smooth process.
Implementing an intelligent speech bubble placement system.
What we learned
Combining multiple AI models efficiently improves results.
Text-to-image generation still struggles with fine details like text placement.
Pre-processing text correctly is crucial for generating accurate visuals.
What's next for Mangify
Improved Character Customization: Allow users to define specific character styles.
Panel Layout Automation: Generate multiple panel layouts dynamically.
Better Action Rendering: Enhance movement depiction within scenes.
User Input for Refinement: Let users refine generated panels with manual tweaks.
Log in or sign up for Devpost to join the conversation.