EggyTalk

EggyTalk ---- A social app that helps you meet new people and communicate with confidence.

Comment

Inspiration

We’ve all experienced moments where we don’t know how to reply in a conversation, especially when talking to someone new. In most social apps, users are expected to already know how to communicate well.  While these platforms focus on matching people, they rarely support users in how to actually talk. At the same time, many AI chat tools generate full responses for users, which can feel unnatural and remove the user’s own voice. Our project is inspired by this gap.
We wanted to create something that feels more human, a system that guides communication instead of replacing it.

What it does

EggyTalk is a social networking app that helps users meet new people and have better conversations through guided support. Instead of automatically generating replies, the app allows users to choose how they want to communicate by selecting different conversation styles. Core features:

  1. Smart Matching System: Users are matched with new people based on their interests and profiles Conversation Mode Selection: Choose how you want to respond (Safe, Relatable, Curious, Unexpected)
  2. AI-Guided Suggestions: The system analyzes both users and suggests ways to continue the conversation
  3. Past Conversation Input: Users can upload previous chats to help the system understand their tone and personality
  4. Personalized Communication Support: Suggestions are tailored to each user’s communication style

How we built it

When users first enter the app, they sign in through a simple welcome page and fill in their personal information. Users can also upload their past conversations with their friends. The purpose is to make AI understand the user’s speech tone. Once matched with a new user, they enter the chat interface. At any point in the conversation, users can choose a conversation mode that is useful when users need help with the conversation topic without embarrassment, which becomes the core interaction of the app. The system then analyzes both users’ profiles and the context of the conversation to provide suggestions that align with the selected communication style. The overall experience is designed to feel lightweight, supportive, and game-like, reducing the pressure often associated with social interaction.

Challenges we ran into

AI Integration & Stability At first, we experimented with different AI providers and ran into issues like rate limits and inconsistent responses. Transitioning to OpenAI required restructuring how we handled requests and ensuring stable, one-click interactions without duplicate calls.

Balancing Guidance vs. Control One of the hardest parts was designing AI guidance that helps without taking over. We had to carefully design the “safe / curious / bold” system so users still feel in control of the conversation instead of relying entirely on AI-generated replies.

Real-time UX Decisions Integrating AI into a live chat interface introduced challenges in timing and responsiveness. We needed to ensure suggestions felt instant and natural, without breaking the conversational flow.

Voice Integration (ElevenLabs) Adding voice seemed simple at first, but required debugging API permissions, handling audio playback correctly, and preventing repeated or global triggers across multiple messages.

Frontend Polish Under Time Pressure With chat UI, dynamic states, icons, and message-level interactions (like per-message audio), keeping the layout clean and consistent within a hackathon timeframe was a challenge.

Accomplishments that we're proud of

Built a fully interactive AI-powered chat experience

Designed a clear communication framework

Maintained user agency

Integrated voice for emotional depth

End-to-end product flow

What we learned

AI is not just about output — it’s about experience design

Small UX decisions matter a lot in chat products

Voice adds emotional context

What's next for EggyTalk

Deeper AI personalization based on long-term user behavior.

Real user-to-user Integration by moving from simulated conversations to actual real-time messaging between users with AI guidance layered in.

Personalized coaching that adapt suggestions based on a user’s past conversations, tone, and communication patterns.

Expanded matching logic for better compatibility between users.

Built With

Share this project:

Updates