VoiceSathi

In India, many people are comfortable speaking their local language but may find it difficult to type long questions, understand complex English content, or navigate multiple websites and applications.

This inspired us to create VoiceSathi, a voice-first AI assistant that allows users to simply speak their questions in their preferred language and receive clear, easy-to-understand responses.

Our goal is to make digital information more accessible to people regardless of their language, typing ability, or technical knowledge. VoiceSathi allows a user to:

Select a preferred language such as Hindi, Telugu, Tamil, Kannada, Marathi, or English. Ask a question using their voice. Convert the voice input into text using speech recognition. Use AI to understand the user's intent and generate a simple response. Read the response back to the user using text-to-speech.

For example, a user could say:

“Mujhe kal Hyderabad jana hai. Main apni journey kaise plan karu?”

VoiceSathi could understand the request and provide simple guidance in the user's preferred language.

The same approach could later be extended to areas such as agriculture, education, government services, travel, and customer support.

How We Built It

We designed VoiceSathi using a simple voice-to-AI-to-voice architecture:

User speaks → Speech-to-Text → AI Processing → Response Generation → Text-to-Speech → User hears response

A prototype can be built with:

Frontend: React or HTML/CSS/JavaScript Backend: Python with FastAPI Speech Recognition: Azure AI Speech AI: Azure OpenAI Text-to-Speech: Azure AI Speech Language support: Azure AI Translator where translation is needed Hosting: Microsoft Azure

The user interface is intentionally simple, with a microphone button, language selector, conversation area, and audio response.

What We Learned

While building VoiceSathi, we learned that accessibility is not only about providing more features. It is also about reducing the effort required to use technology.

We learned about:

Speech-to-text and text-to-speech technologies Generative AI integration Prompt design Multilingual user experiences API integration Designing for accessibility Rapid prototyping Handling errors and unclear voice input

Most importantly, we learned to approach the solution from the user's perspective rather than only from a technical perspective.

Challenges We Faced

One of the biggest challenges was handling differences in pronunciation, accents, background noise, and conversations containing words from multiple languages.

Another challenge was ensuring that AI-generated responses were simple enough for users to understand.

Since hackathons have limited time, we also needed to prioritize the most important functionality rather than attempting to build every possible feature.

We therefore focused our prototype on one strong workflow:

Speak → Understand → Respond → Listen

Future Scope

VoiceSathi could eventually support:

More Indian languages and dialects Agriculture-related assistance Government service navigation Education and learning assistance Travel assistance Voice-based form filling Customer service Low-bandwidth environments Accessibility support for elderly and visually impaired users Impact

VoiceSathi demonstrates how AI and voice technology can reduce digital barriers.

Rather than expecting users to learn complicated interfaces, search using exact keywords, or type lengthy questions, VoiceSathi allows technology to adapt to a natural human behavior:

speaking.

Submission Summary

Project Name: VoiceSathi Category: AI / Accessibility / Voice Technology Target Users: People who prefer speaking over typing, particularly users of Indian languages Problem: Language, literacy, typing, and interface complexity can make digital services difficult to access. Solution: A multilingual voice-first AI assistant that understands spoken questions and provides simple spoken responses. Technology: Azure AI Speech, Azure OpenAI, Azure AI Translator, Python/FastAPI, React Key Innovation: Voice becomes the main interface instead of an additional feature.

One-line pitch

VoiceSathi is a multilingual AI assistant that lets people access digital information naturally by simply speaking in the language they are most comfortable with.