InfoStream

Wins are selective to those who are fast not those who are conventional

Comment

🚀 Inspiration

In today’s fast-paced digital world, people spend more time searching and skimming than actually learning. Students, researchers, and professionals face information overload, struggling to retain complex topics.

InfoStream was created to solve that. It transforms dense text into short, cinematic videos — making learning fast, engaging, and visually intuitive. Our inspiration: democratizing knowledge through AI-driven storytelling.

💡 What it does

InfoStream converts any article, webpage, or block of text into a narrated 60-second video summary using AI reasoning and AWS-powered automation.

It’s not just summarization — it’s visualization. Each video scene is generated intelligently with style options like realistic, cinematic, or anime.

Key Features:

🧠 Amazon Bedrock (Claude 3) — Analyzes text and creates scene scripts.

🎬 JSON2Video / FFmpeg — Generates 8–10 second scene videos.

🖼️ Unsplash / Pexels — Sources contextual visuals.

☁️ Amazon ECS + S3 — Runs jobs and stores final outputs.

🧮 RDS (PostgreSQL) — Tracks video jobs and API usage.

🔐 AWS Secrets Manager — Secures API keys and environment variables.

📊 Amazon CloudWatch — Monitors pipeline metrics and logs.

👉 Input text → AI reasoning→ Scene creation → Video generation → Ready-to-share output.

🧱 How we built it

We designed InfoStream as an agentic AI pipeline, powered by Amazon Bedrock for reasoning and AWS cloud services for orchestration.

1️⃣ User inputs text or URL via frontend.

2️⃣ Backend (Node.js + Express on ECS) creates a job entry.

3️⃣ Amazon Bedrock (Claude 3) breaks text into 5 scenes with scripts.

4️⃣ Image search (Unsplash/Pexels) gathers relevant visuals.

5️⃣ JSON2Video or FFmpeg generates scene videos.

6️⃣ Amazon S3 stores the final stitched video.

7️⃣ Amazon CloudWatch logs and monitors execution.

8️⃣ RDS stores job progress, provider usage, and metrics.

💡 Resilient Architecture Highlights

Circuit breaker and provider rotation ensure zero downtime.

Mock mode allows complete local testing without real API calls.

Secrets Manager ensures secure environment variable storage.

Deployable using Docker + ECS/Fargate or AWS Lambda Containers.

⚙️ Challenges we ran into

Optimizing Claude 3 prompts for meaningful scene breakdowns.

Managing API quotas and failover logic across multiple providers.

Synchronizing visuals and narration at consistent pacing.

Building a resilient fallback system (JSON2Video → FFmpeg).

Handling secure key rotation and IAM policy isolation on AWS.

Each challenge made InfoStream more autonomous, reliable, and production-ready.

🏆 Accomplishments that we're proud of

Built a fully working AWS-integrated AI agent that transforms text to video autonomously.

Achieved sub-60-second video generation time end-to-end.

Integrated failover and circuit-breaker logic for consistent output.

Designed an intelligent provider rotation system.

Implemented mock mode to simulate real AI calls at zero cost.

Deployed a secure, observable pipeline monitored by CloudWatch.

🎯 Result: A complete, scalable, and agentic AI solution built for real-world use.

📚 What we learned

Reasoning LLMs (Claude 3) can autonomously create creative and structured scenes.

ECS/Fargate orchestration simplifies scaling and cost management.

Secrets Manager and IAM roles are essential for production security.

Observability with CloudWatch enables efficient debugging and metrics tracking.

Treating LLMs as decision-making agents rather than passive generators is a game-changer.

We learned how to turn a static AI model into a truly autonomous storytelling engine.

🔮 What's next for InfoStream

🎤 Integrate Amazon Polly for lifelike, multilingual voiceovers.

🤖 Extend reasoning via Amazon Q and Bedrock AgentCore.

📚 Launch an educational API for classrooms and e-learning.

⚙️ Add Amazon Step Functions for more complex orchestration.

🌐 Personalize tone, theme, and pacing for each user.

🔎 Use Amazon Kendra for knowledge-grounded responses.

The goal: make InfoStream a universal knowledge-visualization agent, powered by AWS.

Built With

Share this project:

Updates