We’re excited to introduce NeuroHeart.AI — a generative meditation app built to help users manage stress and anxiety through real-time, voice-guided mindfulness. Inspired by the rise in burnout and emotional fatigue, we envisioned a companion that doesn’t just offer static playlists but generates personalized meditations when they’re needed most.

We built NeuroHeart entirely using Bolt.new, leveraging natural language prompts to create a full-stack application without writing manual frontend code. The journey taught us how to orchestrate multiple systems—Bolt for frontend, Supabase for secure storage, n8n for automation, and ElevenLabs for lifelike audio synthesis.

NeuroHeart.AI taps into the booming \$5.2B global meditation market with a unique, adaptive AI offering. Its real-time emotional personalization positions it as a first mover in AI-driven therapeutic mindfulness solutions.

NeuroHeart’s backend is grounded in meditation-centric published research led by the founders for over 5 years. It operationalizes findings from peer-reviewed studies—including the Chakramarma meditation protocol and HRV-based stress analysis—demonstrating that structured guided meditation can measurably reduce stress and enhance attention. This system was developed through in-person, on-campus research involving real student participants at a university. The meditations followed a structured meditation engine rooted in emotional reflection, challenge analysis, and solution rehearsal. Voice generation is informed by timing, breath pacing, and meditative flow using ElevenLabs meditation voices. The system is built not just on AI algorithms but validated meditation science, bridging ancient contemplative frameworks with reproducible, data-driven outcomes presented at ASEE and FIE conferences.

A major technical challenge was designing and integrating a real-time context retrieval system. Specifically, we needed to build an ingestion pipeline to process and embed 50 research documents, complete with metadata, into a vector database for Retrieval-Augmented Generation (RAG). We overcame this by leveraging n8n to automate the entire flow — from document parsing and metadata extraction to LLM-based summarization and OpenAI embedding — with the results securely stored in Supabase.

The stack includes:

  • Bolt.new: for rapid, fully functional frontend.
  • Supabase with RLS: secure, scalable backend.
  • n8n: automation for embeddings, summarization.
  • OpenAI / Deepseek: dynamic meditation script generation.
  • ElevenLabs: high-fidelity TTS voice generation.
  • Tavus: lip-synced avatar video generation.
  • Vectorized content for smart query-based reflection.

Another significant challenge was building an intuitive, responsive interface that seamlessly supported chat, voice interactions, and user personalization. This was successfully addressed using Bolt.new’s rapid prototyping tools, which enabled us to design and deploy a fully functional frontend without writing manual code.

  • Multi-threaded sessions.
  • Touch-friendly responsive design / mobile friendly.
  • Theme switching.
  • Soft color palettes and no-notification design for mental clarity.

What we learned is that calm can be coded. By combining cognitive-behavioral principles, generative AI, and voice synthesis, we created an experience that feels deeply human. NeuroHeart doesn’t just talk—it listens, reflects, and heals.

The most recent AI-generated message is automatically used as the input script for video generation. ElevenLabs converts this script into lifelike speech, creating a natural voiceover. Tavus then uses this audio to generate a talking-head video, where the avatar precisely lip-syncs to the spoken content, ensuring seamless and realistic visual delivery.

Lip-sync video generation powered by Tavus and ElevenLabs is functional and available for demo here: https://gilded-heliotrope-e2dc52.netlify.app/. While it is not embedded as a core UI element submitted to maintain application stability, the feature demonstrates our immersive meditation. For video generation experimenting, please enter Tavus API key manually on https://gilded-heliotrope-e2dc52.netlify.app/. App Output: Tavus generated video example with lip sync: https://tavus.video/694a86b339.

Unlike static meditation apps like Calm or Headspace, NeuroHeart.AI delivers a dynamic, intelligent experience. By combining Retrieval-Augmented Generation (RAG), real-time script generation, and ElevenLabs TTS, it creates emotionally adaptive meditations tailored to the user’s current state—offering a deeply interactive alternative grounded in conversation, not just content.

Why It’s Unique:

  • Real-time emotional tuning with AI.
  • Audio / Video generation from AI chat not found in competitors.
  • Meditation structure grounded in published, original research (Chakramarma protocol).

We believe NeuroHeart.AI brings deep scientific integrity to emotional well-being. Built from 40 research-rich documents curated by founders with over 15 years’ experience, it offers a compassionate, intelligent companion grounded in rigor, reflection, and transformative intent.

Built With

Share this project:

Updates