Understanding

💡 Inspiration

Conflict is rarely about what we say — it's about how we say it.

We noticed that in emotionally charged moments, people often say things they don’t truly mean, especially in close relationships. Even when the intent is not harmful, tone and wording can escalate situations quickly.

We wanted to build something that doesn't just analyze emotions, but actively helps people communicate better in real time — like a relationship coach that speaks through you.

That’s how Understanding was born:

What if you could hear a calmer, more compassionate version of yourself before things escalate?

🛠️ How We Built It

Understanding is a full-stack AI system that transforms emotional speech into constructive communication:

1. Audio Input

   • Users record or upload speech during moments of frustration

2. Emotion & Intent Understanding

   • Powered by HiggsAudio M3 v3.5
   • Transcribes speech and detects emotional signals + underlying unmet needs

3. NVC-based Rewriting

   • We apply Non-Violent Communication (NVC) principles to rewrite the message
   • Output shifts from blame → observation, reaction → need

4. Voice Cloning Output

   • Using Higgs Audio V2.5 (Eigen AI)
   • The rewritten message is spoken back in the user’s own voice, but calmer and more empathetic

5. Backend System

   • Built with FastAPI for async processing
   • Audio pipeline handled with pydub + VAD for clean segmentation

📚 What We Learned

• Communication is a system, not just words
• AI can do more than generate — it can mediate human relationships
• The combination of:
  • speech understanding
  • structured rewriting (NVC)
  • and voice synthesis
    is surprisingly powerful

Most importantly, we realized:

Sometimes, hearing yourself differently is the first step to becoming better.

Built With

• bosonai
• css
• eigenai
• fastapi
• higgsm3v3.5
• higgsv2.5
• html
• javascript
• python