
AIdeas: Dengbej AI – Kurdish Storytelling with Generative AI
A serverless AI app that summarizes articles, translates them into Kurdish, and generates narrated stories inspired by the Dengbej storytelling tradition.
App Category
Social Impact
My Vision
For centuries, Kurdish communities have preserved their history and collective memory through a powerful oral storytelling tradition known as Dengbej. Dengbejs are traditional storytellers who narrate historical events, cultural experiences, and emotional narratives through a distinctive form of Kurdish singing.
Unlike many societies that relied primarily on written historical records, Kurdish culture has often depended on oral storytelling traditions to preserve cultural knowledge and historical memory. Dengbejs historically acted as storytellers, historians, and cultural narrators within their communities.
Here is an example of a Dengbej performance that illustrates the storytelling tradition that inspired this project:
The vision behind Dengbej AI is to explore how modern artificial intelligence can support and extend this cultural storytelling tradition.
This project transforms written articles and stories into short narrated summaries, while also providing Kurdish translations. The goal is to create a platform where users can listen to stories and information in a format that reflects the traditional storytelling culture.
Users can paste an article or provide a URL, and the system generates:
- English summary
- Kurdish translation
- Audio narration of the summarized story
This allows users to read or listen to summarized information quickly and accessibly.
Why This Matters
The Dengbej storytelling tradition represents a living cultural heritage that has helped Kurdish communities preserve their collective memory for generations.
However, as the world becomes increasingly digital and text-based, traditional oral storytelling formats risk becoming less visible.
At the same time, access to information remains a challenge for many communities who prefer consuming stories through listening rather than reading long written articles.
Dengbej AI addresses two important goals:
1. Cultural Preservation
Highlighting and preserving the Dengbej storytelling tradition by demonstrating how storytelling formats can evolve using modern AI technologies.
2. Information Accessibility
Helping Kurdish speakers stay informed about global news in their own language.
Many Kurdish speakers prefer to listen to information rather than read long articles. However, global news and information are rarely available in Kurdish.
By summarizing articles and translating them into Kurdish, this platform allows users to access global information in a more familiar and accessible way.
Although the current prototype does not yet support Kurdish speech narration, the project demonstrates how AI could eventually support bilingual storytelling experiences.
How I Built This
Dengbej AI uses a serverless architecture on AWS designed to be lightweight, scalable, and easy to deploy.
Architecture Overview

Figure: Serverless architecture of Dengbej AI using AWS Amplify, AWS Lambda, Amazon Bedrock, Amazon Polly, and Amazon S3.
This architecture allows the application to scale automatically while minimizing infrastructure management through fully managed AWS services.
AWS Services Used
Amazon Bedrock
Amazon Bedrock provides access to powerful foundation models.
In this project, the Claude model is used to generate:
- article summaries
- Kurdish translations
Using Bedrock allows the application to perform advanced language processing without managing machine learning infrastructure.
AWS Lambda
AWS Lambda acts as the backend processing layer.
The Lambda function:
- receives requests from the frontend
- extracts article text from URLs
- sends prompts to Amazon Bedrock
- generates narration using Amazon Polly
- uploads audio files to Amazon S3
- returns the response to the frontend
Because Lambda is serverless, the system runs only when users submit requests.
Amazon Polly
Amazon Polly converts text summaries into natural sounding speech.
In the current prototype, Polly generates English audio narration of the summarized story.
Although Polly does not yet support Kurdish speech synthesis, the narrated English summary still allows users to listen to the story.
Future versions may explore multilingual speech technologies that support Kurdish narration.
Amazon S3
Amazon S3 stores the generated audio files.
Each narration is saved as an MP3 file and returned to the frontend through a public S3 URL, allowing the web application to stream the audio.
AWS Amplify
AWS Amplify hosts the web interface of the application.
The web app allows users to:
- paste text
- submit article URLs
- generate summarized stories
- read Kurdish translations
- listen to narrated audio summaries
Amplify also provides automatic deployment whenever new code is pushed to the GitHub repository.
Development Tools
During development, I used Kiro, an AI development environment designed to help builders quickly prototype AI-powered applications.
Kiro assisted with:
- generating initial code structures
- refining Lambda function logic
- improving the frontend interface
- debugging integration between services
Using Kiro accelerated the development process and allowed me to focus on architecture and experimentation with AWS AI services.
Demo
You can try Dengbej AI using the live application.
Live Application
Users can paste text or provide an article URL. The system generates:
- English summary
- Kurdish translation
- Audio narration of the summarized story
Source Code


Challenges and Technical Trade-offs
Supporting Kurdish Narration
One of the biggest challenges was speech synthesis for the Kurdish language.
While Amazon Polly provides high quality neural voices, it does not currently support Kurdish speech synthesis. Because of this limitation, the current prototype generates English narration while still providing Kurdish translations as text.
Future versions could explore multilingual speech models or custom voice technologies that support Kurdish narration.
Extracting Article Content
Another challenge involved extracting clean article content from URLs. Many websites include navigation elements, advertisements, and formatting that can interfere with summarization.
To address this, the backend uses Python libraries such as:
- Requests
- BeautifulSoup
These tools extract paragraph content from HTML pages before sending the text to the AI model.
Balancing Summary Length and Cost
Generative AI models provide powerful summarization capabilities, but longer inputs can increase inference time and cost.
To balance performance and efficiency, the system limits article length before sending content to the AI model.
What I Learned
Building Dengbej AI provided several insights into combining AI with serverless cloud infrastructure.
One key lesson was the power of serverless architectures. By using managed AWS services such as Amazon Bedrock, AWS Lambda, Amazon Polly, and Amazon S3, it was possible to create a fully functional AI workflow without managing servers or machine learning infrastructure.
Another important insight was the challenge of multilingual speech technologies. While AI models can translate text into Kurdish effectively, speech synthesis tools do not yet widely support Kurdish voices.
Finally, this project demonstrated how cloud-based AI services can be used not only for technical innovation but also for cultural preservation and accessibility.
Future Improvements
While the current prototype demonstrates the core idea, several improvements could expand the platform in the future.
Kurdish Speech Narration
One of the most important future improvements would be supporting Kurdish voice narration. This could potentially be achieved through emerging multilingual speech synthesis models or custom-trained voice models.
Improved Article Extraction
Future versions could implement more advanced article extraction techniques to better handle different news websites and formatting structures.
Personalized News Feeds
The platform could allow users to subscribe to topics or sources and automatically generate summarized audio stories.
Infrastructure as Code
Managing the infrastructure using Infrastructure as Code tools such as AWS CDK or Terraform would allow the entire system to be deployed, replicated, or destroyed automatically.
Mobile Experience
A mobile version of the application could provide a more accessible listening experience for users who prefer consuming stories through audio.
Conclusion
Dengbej AI explores how artificial intelligence and cloud technologies can support cultural storytelling while improving access to information.
By combining Amazon Bedrock, AWS Lambda, Amazon Polly, Amazon S3, and AWS Amplify, the project demonstrates how modern AI services can transform written content into accessible bilingual storytelling experiences.
Although this prototype is an early step, it highlights how AI tools can help communities preserve cultural traditions while staying connected to global knowledge.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article