Skip to content

Repository files navigation

HikariStoryBoard

HikariStoryBoard is a GenAI-powered storytelling platform that transforms simple text prompts into cinematic comic-style storyboards with consistent characters, narrative flow, and AI-generated visuals.

Built for the GenAI Zürich Hackathon 2026, this project demonstrates how large language models and diffusion models can work together to create structured visual storytelling.

Inspiration

Storytelling is one of the oldest forms of human expression, yet creating visual narratives like comics or storyboards still requires significant effort and artistic skill.

We asked: What if anyone could turn an idea into a cinematic storyboard instantly?

HikariStoryBoard bridges the gap between imagination and visual creation using Generative AI.

What It Does

  • Converts text prompts into a 4-panel cinematic story
  • Maintains character consistency across panels
  • Generates AI-powered images for each scene
  • Adds dialogue via speech bubbles
  • Produces a complete visual storyboard

Architecture

Untitled design (2)

Tech Stack

Frontend:

  • React + TypeScript (Vite)

Backend:

  • Flask (Vercel serverless)
  • Google Gemini API (story generation)
  • Hugging Face SDXL (image generation)
  • Pillow (image processing)

Model Selection

Story Generation:

  • Gemini 2.5 Flash
  • Fast, structured output, strong narrative quality

Image Generation:

  • Stable Diffusion XL (SDXL)
  • High-quality visuals with good prompt control

Art Style

Cinematic comic style:

  • Semi-realistic
  • Dramatic lighting
  • Consistent characters
  • Graphic novel aesthetic

Example style prompt: cinematic comic illustration, semi-realistic, ultra detailed, dramatic lighting, consistent character design


Challenges We Ran Into

  • Maintaining character consistency
  • Getting structured JSON output from LLMs
  • Handling Vercel serverless limits
  • Fixing frontend-backend deployment issues

Accomplishments

  • Built a full LLM + Diffusion pipeline
  • Achieved consistent storytelling across panels
  • Created an AI-powered storyboard generator

Out-of-the-box: We built a system that understands narrative structure and visual continuity, acting like an AI storyteller and director.


What We Learned

  • Prompt engineering is critical
  • Structured outputs enable multi-step AI systems
  • Deployment issues can break working systems
  • UX matters as much as AI quality

What's Next

  • Multiple art styles (anime, noir, watercolor)
  • Multi-character storytelling
  • Editable panels
  • Export as PDF/comics
  • Faster streaming responses

Setup

Clone: git clone
cd hikaristoryboard

Install: npm install

Run: npm run dev

Frontend: http://localhost:5173

Build: npm run build


Deployment

Frontend: Netlify
Backend: Vercel


API

https://hikaristory-backend.vercel.app/output


Environment Variables

ACCESS_TOKEN=your_huggingface_key
GEMINI_API_KEY=your_gemini_key


Structure

/frontend
/src
hooks/useApi.ts
components/

/backend
/api
index.py

Notes

  • Vercel has execution time limits
  • Ensure correct API route (/output)
  • Redeploy frontend after backend changes

Built With

React
TypeScript
Flask
Google Gemini
Hugging Face
Vercel
Netlify


Impact & SDG Alignment

HikariStoryBoard aligns with the following UN Sustainable Development Goals:

  • Quality Education – Enhances learning through AI-generated visual storytelling
  • Industry, Innovation and Infrastructure – Demonstrates real-world applications of Generative AI
  • Reduced Inequalities – Makes creative storytelling accessible to everyone

Additionally, it contributes to:

  • Decent Work and Economic Growth – Supports creators with faster content generation
  • Responsible Consumption and Production – Encourages efficient digital-first creation workflows

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages