Narrate Mate

Inspiration

As a father of a multilingual child who loves listening to stories in different languages, I realized the need for a smarter reading companion. That’s how Narrate Mate was born — an AI-powered application that reads, translates, and stores knowledge from books and documents.

We can also re-purpose the project with RAG integration, So it can help students prepare for exams, generate practice questions, or assist professionals like lawyers by answering questions directly from complex texts — all in your preferred language, saving time and effort.

What it does

Narrate Mate transforms books (PDFs) into engaging, multilingual audio experiences. It extracts text, generates context, descriptions, and narrates the content like a story — in the language you choose.

How we built it

Narrate Mate is built on a fully serverless, event-driven architecture using AWS services like Lambda, S3, and AI/ML tools. We process PDF files, extract text, and generate context using AI. Terraform manages the entire infrastructure as code, ensuring scalability and consistency.

Challenges we ran into

1. AWS Textract was extracting irrelevant text from images inside the books — such as slogans on T-shirts — adding noise to the knowledge base.
2. Making infrastructure changes was time-consuming and prone to errors without automation.

Accomplishments that we're proud of

1. Integrated Amazon Bedrock with Anthropic Claude-3 to intelligently filter and clean the extracted text, solving the noise problem.
2. Implemented complete Infrastructure as Code using Terraform, making infrastructure changes fast, consistent, and reliable.

What we learned

1. AWS Bedrock - Leveraging foundation models like Claude-3 to improve text quality and context understanding.
2. AWS Textract – Extracting text from complex documents.
3. AWS Polly - Converting text to lifelike speech, enabling natural multilingual narration.
4. AWS Translate - Instantly translating extracted content to multiple languages with reliable accuracy.
5. Terraform - Managing infrastructure as code to automate, scale, and simplify deployments.

What's next for Narrate Mate

We plan to integrate RAG (Retrieval-Augmented Generation) and CAG (Conversational-Augmented Generation) to turn Narrate Mate into a fully interactive learning assistant. Users will be able to ask questions about their books, lessons, or documents and get accurate, AI-generated answers — in their preferred language. We also aim to enhance support for exam preparation, legal research, and multilingual story experiences for kids.

Built With

• amazon-web-services
• apigateway
• bedrock
• cloudfront
• framework7
• lambda
• node.js
• polly
• s3
• sns
• terraform
• textract
• translate