EchoPod

Welcome to EchoPod – Smart Audio Learning on the Go. Your personalized, AI-powered podcast learning experience

Comment

Inspiration

We’re full-time students, drowning in textbooks, staring at screens, and constantly multitasking. But what if learning could happen while living—walking to class, washing dishes, or just lying down with your eyes closed?

We craved a platform that could turn any topic into a personalized podcast, instantly. Not generic content. Not pre-recorded lectures. Something smart, fun, and completely tailored to us. And that's how EchoPod was born—your AI-powered podcast generator, built for the way we actually live and learn.

What it does

EchoPod takes your curiosity and turns it into a binge-worthy audio experience—in under 2 minutes.

Here’s what happens:

  • 🎯 You enter a topic — anything from “Binary Trees” to “Quantum Entanglement”
  • 🧠 Choose a category (Programming, Science, Math, etc.)
  • 🎭 Select a tone (Factual, Conversational, or Storytelling)
  • 📚 Set a difficulty level (Beginner, Intermediate, Advanced)

Then the magic begins 💫

  • EchoPod generates up to 7 well-structured chapters, each with intros, examples, and takeaways — crafted using Amazon Bedrock (Claude Sonnet 3.5)
  • Each chapter is transformed into audio using Amazon Polly, running in parallel for blazing-fast output
  • The final result is a personalized podcast, ready to play:
    • ✅ Synced transcripts
    • 🎧 Chapter navigation
    • 🗣️ Narration modes
    • 🎙️ Speech-to-text support for entering new topics hands-free

It’s your learning assistant, narrator, and podcast host—all in one.

How we built it

EchoPod runs on a fully serverless, event-driven architecture, designed for speed, scalability, and cost efficiency:

🧑‍💻 Frontend

Built with Flutter for smooth cross-platform support (Android & iOS), including gesture navigation and transcript playback.

🧠 Content Generation

  • Uses Amazon Bedrock (Claude 3.5) to generate structured content
  • Custom prompts based on topic, tone, difficulty, and category
  • Outputs a chapter-wise structured JSON saved to S3

🔊 Audio Generation

  • Amazon Polly converts each chapter to MP3
  • Polly tasks run in parallel using AWS Step Functions Map State
  • Audio files are stored in Amazon S3, organized by chapter

📦 Compression & Finalization

  • Audio parts are compressed in parallel using another Step Function
  • A dedicated Lambda function merges and finalizes the podcast-ready MP3

🔐 Authentication

  • Amazon Cognito handles secure login and signup

We built the backend using FastAPI (Python), integrated Boto3 for AWS service access, and managed orchestration with Event Driven Architecture.

Challenges we ran into

  • 💥 Parallel Polly Headaches: Handling multiple Polly tasks while keeping chapter order and combining parts on-the-fly was tricky
  • 🔁 Syncing Audio + Transcript: Polly doesn’t return timestamps, so we had to map our own logic to keep transcript display accurate
  • 🎛️ Prompt Tuning: Getting the perfect tone + difficulty match without making the content dry or too complex took so much tweaking
  • 🔐 Cognito Integration: Handling auth tokens, refreshes, and mobile flow while keeping it smooth for users was a learning curve

Accomplishments that we're proud of

  • 🧠 Full podcast generation in under 2 minutes, with up to 7 chapters
  • ⚡ Parallel processing of both content and audio—speed without sacrificing structure
  • 🎧 Beautiful mobile experience: swipe through chapters, tap to listen, read along
  • 🗣️ Speech-to-text and narration modes make it fun, accessible, and easy to use
  • 💻 Actually used it to revise for a Systems Design exam while cooking pasta. Yes, it worked.

What we learned

  • 🔄 How to design parallel Lambda + Step Function pipelines for high-speed TTS workflows
  • 📦 The power of modular, event-driven backend design
  • 🧠 How prompt design affects everything—tone, quality, engagement
  • 🎯 That people love personalization—especially in how they consume knowledge
  • 🔊 That good UX for audio-first learning is not just possible—it’s essential

What's next for EchoPod

  • 💬 Quiz Mode: AI-generated comprehension checks after each chapter
  • 👥 Narration Style Toggle: Monologue vs Dialogue learning styles
  • 🌍 Multi-language Support: Spanish, Hindi, Korean, Arabic & more
  • 📱 Playlist Builder: Save your favorite podcasts by topic
  • 🔊 Platform Integrations: Spotify, Audible, and LMS tools
  • 🧩 Chapter Visualizer: Auto-generated visuals or diagrams per chapter

Built With

Share this project:

Updates