🧠💬 MindMeld – Your AI-Powered Mental Wellness Companion
Built with Bolt, ElevenLabs, Tavus & Supabase | Submission for the World's Largest Hackathon Writing Challenge
🌟 Inspiration
In today's fast-paced world, mental health support is still inaccessible or stigmatized for many. We envisioned MindMeld as a private, always-available companion—an AI-powered mental wellness assistant that can detect emotions, have reflective conversations, and even respond with a calming voice or personalized video.
Our motivation? Build a truly empathetic experience using multi-modal AI, bridging the gap between human warmth and digital intelligence.
💡 What It Does
MindMeld is a web-based AI-powered chat app that helps users manage emotional wellness. It offers:
- 🎙️ Voice input & natural conversation
- 😌 Real-time emotion detection (from text & voice)
- 🧘♀️ Wellness-focused response generation
- 🔊 AI-generated calming voice responses (via ElevenLabs)
- 📹 Personalized reflection video playback (via Tavus)
- 📊 Session tracking and insights (via Supabase)
Users can simply type or speak how they feel—and MindMeld listens, reflects, and supports.
🧱 How We Built It
We combined a powerful stack of frontend, backend, and AI integrations:
⚡ Sponsor Tech – Bolt
- We used Bolt to power backend workflows—especially managing AI pipeline triggers.
- Bolt’s low-latency execution enabled us to chain NLP + emotion detection + audio/video generation efficiently.
-
Its developer-friendly API helped us manage queues and retries for tasks like:
- Uploading emotional data
- Triggering Tavus reflection video generation
- Audio generation via ElevenLabs
- Supabase logging
🚀 Favorite Bolt Feature: The event-driven function triggers—made it super easy to link emotion analysis → voice generation → storage. It saved us hours of DevOps.
🛠️ Tech Stack
| Technology | Purpose |
|---|---|
| React.js | Frontend interface |
| Supabase | User session logging & emotion tracking |
| ElevenLabs API | Voice cloning & AI voice playback |
| Tavus API | Personalized AI video generation |
| Bolt | Workflow orchestration & trigger-based backend |
| OpenAI GPT-4o | Chat & emotional response generation |
| Web Speech API | Voice input + output for chat |
🚧 Challenges We Faced
- Emotion Analysis Accuracy: Combining text + voice tone to get accurate emotions required experimentation.
- Video Generation Time: Tavus responses took time to generate—we optimized it with Bolt’s queuing + progressive response.
- Cross-modal sync: Syncing voice, text, and video to ensure one coherent user session was tricky.
- Privacy concerns: We ensured no session data is stored without user consent.
🎉 What's Next
- Android & iOS mobile app with push notifications
- Mood journaling and long-term emotional analytics
- More personalized AI video personalities
- Integration with mental health professionals (referral system)
🤝 Team
Built by @garg80884 and powered by passion 💖 for tech + empathy.
🖥️ Live Demo
- 🌐 Live App
🚀 Final Thoughts
MindMeld is more than just a project—it’s our attempt to bring emotional wellness into everyone's hands using accessible AI. Thanks to Bolt, we stitched together an entire mental wellness experience with real-time automation and ease.
Let’s use AI not just to solve problems—but to heal, reflect, and grow. 🌱
Top comments (0)