Inspiration

We wanted to help differently abled people overcome communication barriers by creating a tool that converts speech to text and text to speech in real time.

What it does

This project converts spoken words into written text and reads typed text aloud in real time, all running offline on Arm-based mobile devices. It helps differently abled users communicate by providing speech-to-text and text-to-speech functionality with multilingual support and full privacy.

How we built it

We built VoiceBridge using lightweight speech-to-text and text-to-speech models optimized for Arm architecture. The entire processing runs locally on the device to ensure privacy and offline functionality. We integrated multilingual support and focused on user-friendly workflows.

Challenges we ran into

One of the main challenges was optimizing AI models to run efficiently on limited hardware without sacrificing accuracy. Ensuring offline operation and privacy while maintaining real-time responsiveness required careful engineering and model compression techniques.

Accomplishments that we're proud of

a beginner-friendly, socially impactful project aligned with hackathon goals, it is independent of internet

  • Supported multiple languages including English, Tamil, and Hindi to enhance accessibility.
  • Ensured complete user privacy by processing all data locally We are proud to
  • Developed a fully offline AI-powered speech-to-text and text-to-speech tool optimized for Arm devices.
  • Achieved real-time performance with lightweight models balancing accuracy and efficiency.

What we learned

Throughout the project, we learned about the complexities of speech recognition and synthesis, especially on resource-constrained Arm devices. We gained hands-on experience optimizing AI models for mobile performance and balancing accuracy with efficiency.

What's next for VoiceBridge

Next, we plan to expand language support to more regional and global languages to reach a wider audience. We aim to improve the accuracy and speed of our models using the latest AI advancements while keeping everything offline for privacy. We also want to add customizable voice options for a more personalized experience. Integration with other accessibility tools and platforms is another goal to create a seamless user ecosystem. Finally, we will conduct extensive user testing to gather feedback and continuously enhance usability and performance.

Built With

  • python
  • using-lightweight-ai-frameworks-like-deepspeech-and-espeak-ng-for-speech-recognition-and-synthesis.-it-runs-on-arm-based-mobile-devices
Share this project:

Updates