Centangle
AI AccessibilityBilingual Screen ReaderUrdu Text-to-SpeechAssistive Technology

Rehnuma Awaz — Pakistan’s First Bilingual AI Screen Reader

Rehnuma Awaz — Pakistan’s First Bilingual AI Screen Reader

Centangle developed Rehnuma Awaz as a bilingual AI screen reader designed to support Urdu and English text-to-speech for blind and visually impaired users. Built with offline-first voice assistive technology, natural voices, Windows and Android support, and screen-reader integration, the platform addresses a major local-language accessibility gap for millions of users in Pakistan.

26M+
visually impaired people addressed
5K+
Android downloads on Google Play Store
3,200+
Windows public listing installs
Rehnuma Awaz — Pakistan’s First Bilingual AI Screen Reader

Project Overview

Giving voice to Urdu and English digital content

Rehnuma Awaz was developed to improve digital accessibility for Pakistan’s blind and visually impaired users by providing a bilingual AI screen reader with Urdu and English text-to-speech capability. The platform responds to a national accessibility gap where many existing screen readers are English-centric and provide poor Urdu support, making local-language digital content difficult to access independently for users who rely on assistive technology.

Giving voice to Urdu and English digital content

  • The Mandate

    Build a local-language assistive technology platform

    The requirement was to create a bilingual screen-reader solution that could support Urdu and English speech output across practical user environments. The platform needed to combine Urdu text processing, pronunciation rules, neural text-to-speech, natural voices, offline support, Windows accessibility through NVDA, Android accessibility through TalkBack, and a user experience suitable for blind and visually impaired users who depend on audio navigation.

  • The Environment

    A major accessibility gap for Urdu-speaking blind users

    Pakistan has over 2 million blind people and 24 million visually impaired people facing access challenges, while many digital tools remain difficult to use without reliable assistive technology. Existing English-centric screen readers often provide limited Urdu support, creating language barriers and digital exclusion for Urdu-speaking blind users. This made Rehnuma Awaz important not only as a technology product, but as a local-language accessibility intervention designed around real user needs.

  • Our Role

    AI development, voice technology, and accessibility engineering

    Centangle worked across the design and development of the bilingual screen-reader experience, combining Urdu language processing, neural text-to-speech, offline voice models, Windows NVDA integration, Android TalkBack support, and multi-platform accessibility workflows. The work focused on making Urdu and English digital content easier to hear, navigate, and understand through natural voice assistive technology.

The Challenge

Digital Urdu content was not accessible enough through existing tools

Blind and visually impaired users depend on screen readers to access digital content, but mainstream tools often perform better for English than for Urdu. This created three connected problems: digital exclusion caused by limited assistive technologies, a language barrier caused by poor Urdu support in English-centric screen readers, and a systemic accessibility gap for Urdu-speaking blind users.

Digital Urdu content was not accessible enough through existing tools

  • Digital Exclusion

    Many blind and visually impaired users faced barriers because reliable assistive technologies for local-language content were limited.

  • Language Barrier

    English-centric screen readers provided poor Urdu support, making digital Urdu content harder to access and understand.

  • Systemic Accessibility Gap

    Urdu-speaking blind users were underserved by mainstream digital accessibility tools and needed a locally relevant solution.

  • Multi-Platform Need

    Users required access across practical environments, including Windows desktop workflows and Android mobile usage.

The Solution

Bilingual AI Access

Centangle developed Rehnuma Awaz as Pakistan’s first bilingual AI screen reader, combining Urdu and English text-to-speech support, natural voices, multi-platform access, offline-first voice models, NVDA integration for Windows, and TalkBack support for Android. Built for a country where over 2 million blind people and 24 million visually impaired people face access challenges, the solution enables users to access digital content through clearer local-language audio across desktop and mobile environments.

The System Overview

  1. Presentation Layer

    Accessible Experience

    • Screen Reader Interface

      Provides users with spoken Urdu and English content through Windows NVDA and Android TalkBack accessibility environments.

  2. APPLICATION LAYER

    • Text Processing Engine

      Handles Urdu and English text input, Urdu normalisation, pronunciation rules, language processing, and preparation for speech generation.

    • Neural TTS Pipeline

      Converts processed text into audio using a text to acoustic model to vocoder pipeline.

    • Windows and Android Connectors

      Supports NVDA add-on integration on Windows and native Android integration with TalkBack support.

  3. DATA & INTEGRATION LAYER

    • Offline Voice Models

      Maintains downloadable and versioned voice models for offline text-to-speech use.

    • Curated Urdu Speech Dataset

      Supports model training, pronunciation handling, Urdu speech quality, and language-specific voice performance.

    • Local Communication Layer

      Uses gRPC for local service communication between the speech engine and platform components.

    • Low-Latency Processing

      Optimises speech generation for responsive and offline assistive technology use.

  • FEATURE 01

    Bilingual Urdu and English Support

    The platform supports both Urdu and English text-to-speech, helping users access mixed-language digital content more effectively.

  • FEATURE 02

    Natural Voice Output

    Rehnuma Awaz provides natural voices to make listening smoother, clearer, and easier for users who depend on audio.

  • FEATURE 03

    Windows NVDA Add-on

    The Windows implementation includes an NVDA add-on developed with Python, enabling screen-reader support within a familiar desktop accessibility environment.

  • FEATURE 04

    Android TalkBack Support

    The Android implementation provides native mobile accessibility through TalkBack support, helping users access content on smartphones.

  • FEATURE 05

    Offline Voice Models

    The platform supports offline and versioned voice models, reducing dependency on constant internet access for speech output.

  • FEATURE 06

    Neural TTS Pipeline

    The system uses a neural text-to-speech pipeline that moves from text processing to acoustic modelling and vocoder-based audio generation.

Delivery Approach

Structured delivery from language research to multi-platform accessibility

The project followed a technology-led accessibility delivery process focused on Urdu language handling, neural speech generation, platform integration, and user-ready assistive access.

  1. 01

    PHASE ONE

    Accessibility and Language Understanding

    Understanding the needs of blind and visually impaired users, the limitations of existing screen readers, and the accessibility challenges faced by Urdu-speaking users.

    • Accessibility requirements
    • Urdu language context
    • User need definition
    • Assistive workflow direction
  2. 02

    PHASE TWO

    Urdu Text Processing and Voice Model Planning

    Structuring Urdu text normalisation, pronunciation rules, speech dataset requirements, and voice model planning for text-to-speech generation.

    • Urdu text-processing rules
    • Pronunciation handling
    • Speech dataset direction
    • Voice model plan
  3. 03

    PHASE THREE

    Neural TTS Development

    Developing the neural text-to-speech pipeline using text processing, acoustic modelling, vocoder-based generation, and low-latency optimisation.

    • Neural TTS pipeline
    • Speech synthesis workflow
    • Voice output models
    • Performance optimisation
  4. 04

    PHASE FOUR

    Windows and Android Integration

    Building the Windows NVDA add-on and Android TalkBack-supported experience with offline model support and platform-specific accessibility integration.

    • Windows NVDA add-on
    • Android accessibility support
    • Offline voice models
    • Platform integration
  5. 05

    PHASE FIVE

    Testing, Refinement, and Public Access

    Testing speech output, platform behaviour, offline performance, user experience, and deployment readiness across Android and Windows access points.

    • Speech validation
    • Accessibility testing
    • Performance refinement
    • Public access readiness
Measured Impact

What changed after Rehnuma Awaz was developed

Rehnuma Awaz made Urdu and English digital content more accessible through bilingual voice support, offline speech models, and screen-reader integration across Windows and Android.

26M+
Built for Pakistan’s wider visually impaired community, including over 2 million blind people and 24 million visually impaired people who face digital access barriePeople Addressedrs.
People Addressed
5K+
Reflecting growing mobile adoption through the Rehnuma Awaz Android app on the Google Play Store.
Android Downloads
3,200+
Showing desktop accessibility reach through the Windows public listing and NVDA-supported environment.
Windows Installs
90%
Indicating strong reported speech performance and positive user response for the bilingual Urdu and English accessibility experience.
Accuracy with 4.8 Rating

Before

  • Urdu-speaking blind users had limited access to reliable local-language screen-reader support.
  • English-centric screen readers provided poor Urdu output.
  • Digital content remained difficult to access without visual reading.
  • Users faced language barriers across everyday digital environments.
  • Assistive technology lacked sufficient offline-first Urdu support.

After

  • Urdu and English content could be accessed through bilingual speech support.
  • Users gained Windows and Android screen-reader access through NVDA and TalkBack integration.
  • Offline voice models reduced dependence on constant internet connectivity.
  • Natural voices improved the listening experience.
  • The platform recorded adoption across Android downloads, Windows installs, accuracy, and user ratings.

Work With Us

Need to make digital content more accessible through AI voice technology?

Centangle helps organisations build AI and assistive technologies that improve accessibility, language inclusion, and digital independence.