Emotional C3PO

Image generated (describing the confusion of Abdelrahman in Today's hackathon)
The app sends reminders to your safety network (eg; friends)
Next.js static page displaying the pipeline of our workflow visually

🌱 Inspiration

We were inspired by a simple yet powerful question: What if your inner voice had a visual language?

Too often, when people feel foggy, overwhelmed, or emotional, they struggle to name or explain what’s going on. We built Nura to help people turn those fuzzy thoughts into clarity—using voice, metaphor, and visuals.

We imagined a space where emotional reflections feel natural, safe, and vivid—and where users leave with something they can see and use again later.

🛠️ What It Does

Nura is an emotionally attuned, multilingual AI assistant that:

Listens deeply to your spoken thoughts
Extracts insights like emotions, metaphors, and internal “scenes”
Draws them as visual metaphors (via Vizcom, Notion, or Slack)
Offers next steps—from mantras to action plans or letting go

It’s your inner voice—but more articulate, more compassionate, and finally… visual.

🔧 How We Built It

Voice-to-AI: We used Vapi.ai for real-time voice capture and assistant interaction
LLM logic: A custom prompt pipeline using OpenAI to guide Nura’s emotional reflection journey
Visualization: Insights are transformed into drawings via Vizcom or written summaries to Notion
Backend: A custom Next.js API endpoint handles Vapi webhook events and processes transcript data
Tagging System: Metadata (emotion, shape, color, temperature, etc.) is embedded for downstream use

🧠 What We Learned

How to design emotional UX for voice-based AI
How to turn abstract concepts (like mood or metaphor) into concrete data
How to build multi-modal pipelines across audio, LLM, and generative visual APIs
How to listen—not just technically, but empathetically

⚠️ Challenges We Faced

Integrating real-time voice with LLMs while maintaining emotional pacing
Creating a prompt architecture that feels human, not scripted
Handling custom workflows using OpenAI without native Vapi flow tools
Mapping messy human feelings → JSON-friendly structured data → generative visuals