Speak in Hindi. Get answers in Hindi. No typing, no English, no barriers.
Millions of Hindi-speaking Indians — farmers, students, daily wage workers — struggle to access basic information because most digital tools are English-first and text-first. Someone who wants to know about a government scheme like PM-KISAN has to navigate English websites, forms, and menus they cannot read comfortably. There is no simple voice-first way for them to just ask and get an answer in their own language.
Themes Selected:
- Human Experience & Productivity
- Climate & Sustainability Systems
- HealthTech & Bio Platforms
- Learning & Knowledge Systems
- Work, Finance & Digital Economy
- Infrastructure, Mobility & Smart Systems
- Trust, Identity & Security
- Media, Social & Interactive Platforms
- Public Systems, Governance and Civic Tech
- Developer Tools & Software Infrastructure
This app lets a user speak a question in Hindi and receive a spoken answer in Hindi — no reading, no typing required.
- Target users: Hindi-speaking individuals in India who are more comfortable speaking than typing or reading English — farmers, students, daily wage workers, and elderly users.
- The pain point: Government schemes, everyday information, and digital services are locked behind English text interfaces that exclude a huge portion of India's population.
- The value: A completely voice-first interaction — speak your question, hear your answer, in Hindi, in seconds.
victors
- Abhishek Kumar (GitHub: abhishek220205-hash) — Backend + AI Integration + Project Lead
- Jigyanshu (GitHub / LinkedIn) — Frontend / Expo Development
- ABHAY (GitHub / LinkedIn) — Docs, PPT,
- Sahil Dogra (GitHub / LinkedIn) — Demo Video, Submission
- We chose this problem because voice access for Hindi speakers is a real, everyday gap — not a hypothetical one.
- The biggest technical challenge was chaining three separate AI API calls (Speech-to-Text → Chat → Text-to-Speech) reliably in sequence, handling markdown-formatted AI replies that broke the Text-to-Speech API, and working around a 500-character input limit on TTS.
- We deliberately kept the app scope to exactly 3 screens — Home, Listening, Result — to make sure the core voice loop was rock solid rather than spreading effort across unnecessary features.
- Frontend: Expo (React Native), SDK 54, JavaScript
- Backend: Node.js, Express
- Database: None (stateless voice pipeline)
- APIs: Sarvam AI — Speech-to-Text (saarika:v2.5), Chat Completion (sarvam-30b), Text-to-Speech (speaker: vidya)
- Hosting: Render (backend deployment)
- AI / ML
- Web3 / Blockchain
- Cyber Security
- Cloud
- Expo Track – Built entirely using Expo (React Native), core functionality runs within the Expo ecosystem, tested live via Expo Go on a real Android device.
- Neo4j Track
- Base44 Track
Sarvam AI Track:
Sarvam AI is the core of this product, not an add-on. All three Sarvam APIs are used in a live, sequential pipeline: Speech-to-Text converts the user's Hindi voice into text, Chat Completion generates a Hindi answer, and Text-to-Speech converts that answer back into spoken audio. Without Sarvam's APIs, the app has no functionality — this is the entire product.
Expo Track:
The app is built end-to-end using Expo (React Native, SDK 54). All three screens (Home, Listening, Result) run natively within the Expo ecosystem, and audio recording/playback uses expo-av. The app was tested live on a physical Android device via Expo Go throughout development.
- ✅ One-tap voice recording — no forms, no typing
- ✅ Real-time Hindi speech-to-text transcription via Sarvam STT
- ✅ Context-aware Hindi answers via Sarvam Chat (sarvam-30b)
- ✅ Natural-sounding spoken responses via Sarvam TTS (auto-plays on result screen)
- ✅ Clean 3-screen flow: Home → Listening → Result
- ✅ Fully functional on a real Android device via Expo Go
- Demo Video Link (Mandatory): [Paste link]
- Deployment Link (Recommended): [Paste Render backend URL]
- Pitch Deck / PPT (Optional): [Paste Gamma.app link]
- All team members completed the mandatory social task
- Bonus Task 1 – Badge sharing
- Bonus Task 2 – Blog/article
- Node.js (v18 or higher)
- Expo Go app installed on a physical Android/iOS device
- A Sarvam AI API key (get one at dashboard.sarvam.ai)
# Clone the repository
git clone https://github.com/abhishek220205-hash/HackHazards-Voice-App.git
cd HackHazards-Voice-App
# ---- Backend ----
cd backend
npm install
# Create a .env file with:
# SARVAM_API_KEY=your_api_key_here
node server.js
# Server runs on http://localhost:3000
# ---- Frontend ----
cd ../frontend
npm install
npx expo start
# Scan the QR code with Expo Go on your phoneNote: The frontend's App.js currently points to a local IP address (http://192.168.31.245:3000) for the backend URL. Update this to your deployed Render URL, or your own local IP (find it via ipconfig), before running.
- 📈 Support for more Indian languages beyond Hindi (Tamil, Telugu, Bengali, etc.)
- 🛡️ Offline fallback mode for low-connectivity areas
- 🌐 Expand beyond government schemes to healthcare, education, and local services information
- Sarvam AI — Speech-to-Text, Chat, and Text-to-Speech APIs
- Expo — React Native framework
- Built as part of HackHazards 2026 (NAMESPACE Community)
Built by a first-time hackathon team from scratch — starting with zero prior experience with Expo, Node.js backend servers, or the Sarvam AI APIs. The biggest breakthrough moment was getting the full voice loop (speak → understand → answer → speak back) working end-to-end on a real phone for the first time. What started as a test script grew into a fully working mobile app within days.