Vaani – Voice Translation App

Vaani: Information safety
Vaani: It learns with time and usage
Offline mode for low connectivity areas
"Emergency Mode: Offline access to critical phrases for urgent situations and low-connectivity areas."
"Vaani Home Screen: Quick access to translation modes, language selection, and emergency features."
"Conversation Mode: Seamless two-way communication between speakers of different languages."
Real-time Voice Translation: Document translation
"Vaani Home screen with toggle for online and offline"
"Vaani Home screen with online mode on"
Text to speech translation

Inspiration

India is a nation of over 1.4 billion people with extraordinary linguistic and cultural diversity. With 22 official languages and hundreds of dialects, language barriers affect daily communication, emergency responses, education, tourism, and inter-state business.

We wanted to build something that truly captures the spirit of "Unity in Diversity"—helping people communicate across languages while still respecting cultural context.

Vaani was inspired by this need. It also serves foreign tourists who want to explore India but struggle with local languages.

What it does

Vaani is a real-time voice and text translation app designed for India.

Key Features:

Real-time voice-to-voice translation between Indian languages and global languages
Document translation from images (OCR-based)
Emergency Mode with offline critical phrases
AI voice learning that adapts to user pronunciation
Accessibility support for visually and hearing-impaired users
Offline functionality for use in low-connectivity regions
Cultural context preservation, ensuring idiomatic and region-appropriate translations

How we built it

Built entirely on Bolt.new
Frontend: React with Tailwind CSS, shadcn/ui, and Framer Motion for UI/UX
Backend Simulation: Node.js with Express
APIs Used:
- WebSpeech API for speech recognition and text-to-speech
- Google Translate API for language translation
Database: IndexedDB for offline storage
Deployment: Netlify

We also designed specialized modes like Tourist Mode, Emergency Mode, and Document OCR Translation, making the app versatile for real-world use cases.

Challenges we ran into

Multilingual dialect handling: India’s regional dialects made translation accuracy and pronunciation difficult to manage.
API limits: Balancing free API quota while ensuring real-time performance was tricky.
UI accessibility: Designing features that work for visually impaired and non-verbal users required extra testing.
Submission hurdles: We encountered platform-level issues while submitting (Devpost account suspension, Bolt export challenges, etc.), but we documented and persisted.

Accomplishments that we're proud of

Built and deployed a fully functional, multi-language voice translator for India within the hackathon timeframe.
Successfully integrated speech recognition, TTS, and real-time translation pipelines.
Developed offline emergency features for critical situations.
Added accessibility layers for differently-abled users.
Created a high-quality demo video and presentation, entirely AI-narrated.

What we learned

How to rapidly prototype and deploy production-ready apps on Bolt.new
Working with real-time speech APIs
Addressing multilingual user experience challenges
Building apps that are both functional and inclusive
Importance of submission readiness and documentation when platform issues arise.