Res-Q

Real-Time Multilingual Disaster/Emergency Response Assistant

Comment

Inspiration

During emergencies, every second counts. We wanted to create an AI-powered system that allows people to communicate in any language, even in high-stress or physically restrictive situations. Whether someone’s trapped or multitasking, Res-Q makes it possible to get help or access critical info using only your voice.

Our goal was to combine real-time speech interaction, AI reasoning with multilingual capabilities, and natural voice synthesis to build a tool that’s both practical and human-friendly, not meant to replace police or services but rather to supplement them.

What it does

Res-Q is a voice-based AI assistant that enables natural, conversational interaction with an emergency response system. With just your voice, you can:

  • Ask for help or describe a situation
  • Get real-time info scraped from trusted sources (e.g., weather alerts, nearby hospitals)
  • Automatically detect your approximate location using IP-based geolocation
  • Hear natural, lifelike voice responses generated by ElevenLabs
  • Have intelligent, context-aware conversations powered by Google Gemini AI

And the best part is, you can do this in any language!

How we built it

Res-Q is built on top of the Gemini Voice Base Stack, integrating:

  • SpeechRecognition + PyAudio for continuous microphone listening
  • Google Gemini API for text generation and reasoning
  • ElevenLabs API for real-time text-to-speech (TTS)
  • IP-based geolocation for automatic context detection
  • Web scraping to pull in real-time information during conversations
  • Python backend with a React.js frontend

Challenges we ran into

  • Integrating frontend and backend
  • Multilingual aspect
  • Managing real-time audio streaming without lag
  • Syncing speech detection and TTS playback
  • Handling API latency for smooth interaction

Accomplishments that we're proud of

  • Built a fully functional real-time voice AI stack from scratch
  • Achieved smooth voice interaction with Gemini and ElevenLabs
  • Implemented auto-pause detection and interruptible playback
  • Designed the foundation for a rescue-focused AI assistant

What we learned

  • How to integrate LLMs with audio interfaces
  • Optimizing real-time speech recognition pipelines
  • Handling multi-threaded playback and recording in Python
  • Designing for hands-free accessibility and inclusivity

What's next for Res-Q

  • Launch a web dashboard for monitoring and control
  • Create a mobile version for on-the-go use
  • Make more flexible with different languages
Share this project:

Updates