Inspiration
SafeSpeak was inspired by the reality that many women do not feel safe in everyday environments–whether it’s walking home at night, riding in an Uber, waiting for public transit, or navigating unfamiliar spaces alone. In these moments, fear often arises subtly and suddenly, long before a situation escalates into an emergency.
Our goal is to create a discreet tool that empowers women to protect themselves. By enabling voice-triggered protection, audio recording, and socially acceptable exit mechanisms, SafeSpeak supports women in environments where safety should be assumed, but often isn’t.
Users can activate SafeSpeak with a simple press of a button, and start simulating a call. A call simulation provides non-confrontational ways for users to exit unsafe or uncomfortable situations, addressing early-stage risks that may not always result in danger, but are avoidable nonetheless.
What it does
- SafeSpeak: One-tap activation to prepare the device for safety monitoring
- Incident Active State: Time starts, automatically records audio when triggered.
- Fake Call Simulation: Using AI text to audio conversion to create a realistic human voice and conversation and warn off potential danger.
How we built it
- Frontend: React Native + Expo, Typescript, Tailwind, CSS
- Backend: Javascript, Node.js, Supabase, Gemini, ElevenLabs
- Designed the User Flow and Interface
- Built the Frontend with React Native + Expo
- Integrated Audio Recording and Haptic Feedback
- Implemented Keyword-Based Trigger Detection
- Developed Fake Call Simulation with AI Voice
- Connected Backend Services with Node.js and Supabase
- Enabled Notifications and Safety Alerts
- Tested for Real-World Cases
Challenges we ran into
- Integrating ElevenLabs Text-to-Speech API
- Balancing iOS Platform Constraints
Accomplishments that we're proud of
- Built within time constraint, including front-end UI, audio recording, and fake call simulation
- Learned to integrate ElevenLabs for text-to-speech and speech-to-text
- Implement Gemini API
What we learned
- Integrated AI-driven voice and dialogue systems
- Applied generative AI to realistic safety use cases
- Coordinated frontend, backend, and APIs effectively
What's next for safespeak
- Expand AI-based detection for situational triggers such as motion patterns, background audio
- Conduct user testing to refine UX for emergency and non-emergency scenarios.
Built With
- elevenlabs
- gemini
- javascript
- node.js
- react
- supabase
- typescript
Log in or sign up for Devpost to join the conversation.