DreamWeaver AI

AI-powered micro-TV showrunner that brings stories to life.

Comment

Inspiration

We wanted to reimagine storytelling with GenAI — not just through text or images, but through entire stylized episodes. Our goal was to create a tool that feels like your own personal showrunner, where prompts turn into animated scenes, voices, and stories. With generative models rapidly advancing, we knew the time was right to build something that blends creativity, interactivity, and AI into a fun, compelling experience.

What it does

DreamWeaver AI transforms any text prompt into a short, stylized micro-TV episode using multi-modal GenAI. It:

  • Takes a user prompt + genre + duration
  • Uses GenAI to generate story scenes, voiceover, and imagery
  • Assembles it into a preview-ready mini-episode with scene cards and simulated audio/video
  • Lets users create again, explore different genres, and visually experience storytelling

How we built it

We built DreamWeaver AI as a React-based frontend using:

  • Tailwind CSS + Framer Motion for a stunning animated UI
  • Amazon Bedrock for LLM-based story & scene generation
  • Amazon SageMaker JumpStart and Titan Image APIs for generating stylized scene images
  • Amazon Polly for voice synthesis (stubbed for now)
  • Remotion and FFmpeg planning for future video assembly

We did not use a separate backend — we directly integrated AWS SDKs and AI services into the React frontend using official AWS React libraries. This kept our architecture serverless, secure, and lightweight.

Challenges we ran into

  • Creating smooth, cinematic UX with limited rendering time constraints
  • Aligning story-to-image transitions while using mock data for demo purposes
  • Building a dynamic multi-modal pipeline without a backend in just days
  • Deploying using serverless AWS services and managing API integrations from React only

🏆 Accomplishments that we're proud of

  • Designed a fully animated, responsive, production-ready UI
  • Delivered a working end-to-end user journey with mock episodes
  • Integrated Bedrock-based LLMs directly into React
  • Created a creative, original GenAI product that's both fun and technically robust

📚 What we learned

  • How to orchestrate multi-agent GenAI workflows without a backend
  • How to optimize GenAI latency for better UI experience
  • Deepened our hands-on knowledge of Amazon Bedrock, Titan, Polly, and Sagemaker JumpStart
  • Best practices for deploying GenAI apps using AWS and React together

🚀 What's next for DreamWeaver AI

  • Integrate full backend rendering with Remotion and FFmpeg to auto-generate full video episodes
  • Add support for user avatars, multi-character stories, and custom voices
  • Enable public sharing and community content gallery
  • Extend to education, mental wellness, and personalized entertainment applications
  • Fine-tune LLM prompts to better structure scenes and character arcs
Share this project:

Updates