Zenith

Turn real-world news into kid-friendly cartoon stories in minutes.

Zenith is a hackathon project built for HackTheEast 2026 that converts a complex topic or news URL into a short, multi-scene cartoon video. By transforming adult news and complex topics into engaging, age-appropriate animated content, Zenith provides a safe and educational digital environment for children.

🏆 Hackathon Tracks

Zenith is designed with the following HackTheEast 2026 tracks in mind:

• Ingram Micro & AWS Agentic AI Champion: Zenith utilizes AWS Bedrock (Amazon Nova Pro) to power its core agentic text generation capabilities. It intelligently structures raw news into engaging story scripts, character profiles, and detailed scene prompts, bringing the agentic workflow to life.
• MiniMax Creative Usage Award: Zenith heavily utilizes the MiniMax API suite (Video, Image, and TTS) to power a highly creative, multimodal AI experience. It orchestrates multiple MiniMax tools to generate characters, scene images, and ultimately links them together into a cohesive video story.
• ExpressVPN Digital Guardian Award: Zenith focuses on children's digital well-being by transforming the overwhelming and often inappropriate internet news landscape into safe, educational, and easy-to-understand cartoon stories. It's a privacy-first approach to content consumption for kids.
• OAX Foundation AI EdTech Platform Award: Zenith tackles content overload in fast-moving fields by intelligently curating real-world news and complex topics, transforming them into easily digestible, engaging animated stories that accelerate learning.
• RevisionDojo Future of Learning Award: Zenith transforms education by personalizing complex, dry subjects into engaging, multimodal stories, making it significantly easier for younger or struggling learners to grasp difficult real-world concepts through adaptive, impactful content.

🚀 How It Works (The Workflow)

When a user enters a URL of a news page or pastes a topic, Zenith executes the following automated pipeline:

1. Ingestion: We scrape the content of the provided URL using Exa (or if a topic is provided, we search for and retrieve relevant articles).
2. Story Planning: We consolidate the gathered information using AWS Bedrock (Amazon Nova Pro) to create a kid-friendly story plan. This includes defining scenes, writing a script, creating characters, and determining the music style.
3. Prompt Generation: We use AWS Bedrock to generate a set of detailed image generation prompts for the starting and ending frames of each scene.
4. Video Generation: We generate videos using MiniMax Video Generation, utilizing the starting and ending frames to smoothly link the scenes together.
5. Character Dialogue: We use MiniMax Text-to-Speech (TTS) to generate character dialogue based on the script.
6. Playback: The user can play the full, sequential story directly in the app.

✨ Features

• Authenticated users can create workflow tasks from the header dialog.
• URL/topic ingestion with Exa (getContents + searchAndContents) and paywall fallback search.
• Structured task pipeline persisted in Convex with stage/status tracking.
• Story planning and prompt-pack generation using AWS Bedrock text generation.
• Prompt-pack generation with JSON repair fallback.
• Scene image generation from start/end frame prompts using MiniMax.
• Scene video generation with first/last-frame chaining and polling using MiniMax.
• Task detail page with:
  • Source, story plan, prompt pack, image/video asset sections
  • Stage-aware next action button
  • "Play Full Story" sequential playback experience

🛠️ Tech Stack

• Frontend: Next.js (App Router) + TypeScript, Tailwind CSS, shadcn/ui
• Backend & Database: Convex + Convex Auth
• AI Text Generation: AWS Bedrock (Amazon Nova Pro API)
• AI Media Generation: MiniMax (Image, Video, Speech APIs)
• Search & Scraping: Exa API
• Package Manager: Bun

📂 Key Files

• convex/schema.ts - workflow task schema and status model
• convex/workflowTasks.ts - workflow API re-export barrel
• convex/workflowTasks.actions.ts - orchestration actions
• convex/workflowTasks.mutations.ts - task state persistence mutations
• convex/workflowTasks.queries.ts - task queries
• convex/workflow/helpers.ts - normalization/parsing helpers
• convex/workflow/validators.ts - Convex validators
• components/workflow/CreateTaskDialog.tsx - quick task launcher
• app/api/bedrock/route.ts - AWS Bedrock API route
• app/(protected)/tasks/[taskId]/page.tsx - task details + full story playback

🚦 Getting Started

Prerequisites

• Bun
• Convex account/deployment
• AWS Account with Bedrock access
• Exa API key
• MiniMax API key

Setup

bun install
bunx convex dev
bun run dev

Environment Variables

Create .env.local:

CONVEX_DEPLOYMENT=your-deployment
NEXT_PUBLIC_CONVEX_URL=https://your-deployment.convex.cloud

EXA_API_KEY=your-exa-api-key

# AWS Bedrock
AWS_ACCESS_KEY_ID=your-aws-access-key
AWS_SECRET_ACCESS_KEY=your-aws-secret-key
AWS_REGION=us-east-1

# MiniMax
MINIMAX_API_KEY=your-minimax-api-key
MINIMAX_GROUP_ID=your-group-id

🚧 In Progress / Next Steps

• Background music generation + timeline mixing
• Final composed export (single merged output file)
• Optional style presets per task

📄 License

MIT