PeaceMaker

Inspiration

As an international student, with the advent of technology, it's easier than ever to reach our friends and family back home, but it's still hard to stay not only communicating but also genuinely connected. Conversations often drift into shallow small talk, become unbalanced (one person dominates or checks out), or escalate over minor misunderstandings. Over time, this reduces call frequency, relationship quality, and well-being.

What it does

PeaceMaker is a real-time conversation coach for video calls. It analyses speaker volume, tone, and live transcription to gauge sentiment, balance, and tension. When a call starts drifting—e.g., one person is monologuing, interruptions spike, or the mood turns negative—PeaceMaker surfaces subtle, actionable nudges like: • “Try an open question.” • “Acknowledge their feeling.” • “Slow down; give them space.” Prompts are lightweight and non-disruptive, so users can course-correct without derailing the call.

How we built it

• Video & audio: FastRTC calling with a React frontend. • Transcription: Streaming ASR via HuggingFace with speaker diarization. • Signals: Prosody features (volume, pitch, speaking rate, overlap/interruptions) from the audio stream + rolling-window text sentiment from the transcript. • Fusion logic: Rule-based turn-taking heuristics combined with a lightweight classifier and a small transformer to trigger context-aware tips. • UI: On-call coach as subtle toasts and a sentiment indicator. • Privacy: All processing is session-based; no audio is stored.

Challenges we ran into

• Keeping transcription + analysis latency low enough for on-the-fly coaching. • Accurate speaker diarization during overlap and cross-talk. • Designing prompts that help without distracting the user. • Handling cultural nuance and multilingual slang in sentiment. • Balancing “useful signal” vs. alert fatigue.

Accomplishments that we're proud of

This is the first time all 4 of us are participating in a Hackathon, and although we are all from the faculty of IT, none of us has an IT background, nor do we have any startup background. The fact that we successfully came up with a marketable idea and built a working prototype (MVP) is nothing short of a miracle.

What we learned

In this hackathon, we learned how to build a video chatting app with real-time transcription and sentiment analysis. We also learn how to work better in a team and how to use our personal experience to come up with some potentially marketable ideas, and execute the idea, and maybe one day we can be entrepreneurs too.

What's next for PeaceMaker

We will use this product with our friends and families back home to further refine this product, and if all goes well, we also want to share it with the broader international student community to gain more feedback and to perfect this product further.