WalkingPal

• 

  Starting Screen of our web app

Inspiration

Walking home alone can feel unsettling, especially at night or in unfamiliar areas. We wanted to create something that could provide immediate reassurance in those moments, not just through text, but through a familiar, human-sounding voice. The idea for WalkingPal came from the realization that hearing someone you trust can make a situation feel safer, even if that support is virtual.

What it does

WalkingPal is a voice-first AI companion that users can talk to when they feel unsafe or anxious while walking alone. Users can switch between two personas — a Friend or a Family Member — and communicate either by typing or speaking. The agent responds with both text and realistic voice audio, creating a more natural and comforting interaction in real time.

How we built it

We built WalkingPal using a full-stack web approach:

React + JavaScript for the frontend interface

Node.js + Express for backend API handling

ElevenLabs Agents for conversational intelligence and voice generation

Speech Recognition APIs for voice input

Custom CSS styling to ensure the interface is responsive and centered on different screen sizes

The backend routes user input to the appropriate AI agent, retrieves a text response, converts it into speech, and returns both the audio and text to the frontend for display and playback.

Challenges we ran into

We faced several technical challenges during development:

Debugging API key permissions and enabling correct agent access

Handling differences between raw audio and JSON-based API responses

Synchronizing text responses with voice playback

Managing latency while keeping the interaction feeling natural

Adjusting UI layout and styling across various screen sizes

Each challenge required iteration, testing, and a deeper understanding of the tools we were using.

Accomplishments that we're proud of

Successfully integrating real-time AI agents with voice responses

Creating a multimodal experience that combines speech, text, and UI feedback

Designing distinct personas with different tones and behaviors

Building a polished, responsive interface suitable for live demos

Turning a safety-focused idea into a working prototype in a short time

What we learned

Through this project, we learned:

How to work with AI voice and agent APIs

The importance of prompt design in shaping AI behavior

How to debug asynchronous frontend–backend workflows

Best practices for handling audio playback in browsers

How small UX decisions can significantly impact how safe and supported a user feels

What's next for WalkingPal

In future versions of WalkingPal, we plan to:

Add short-term conversation memory for more contextual responses

Improve latency through streaming audio playback

Allow users to customize voices or upload voices of trusted contacts

Introduce optional safety features like emergency contact escalation

Expand accessibility features for hands-free, on-the-go use

Built With

• api
• css
• elevenlabs
• html
• javascript
• node.js
• react