Completed

ZenGen

AI-guided meditation with custom visuals and voiceovers

System Architecture Visual

Project Overview

ZenGen generates fully personalised meditation sessions on demand. Users describe their intention or mood, and the app uses an LLM to craft a bespoke guided script, a text-to-speech engine to narrate it, and an AI image generator to produce unique ambient visuals — delivering a cohesive, immersive meditation experience.

Methodology & Architecture

1. Session Scripting

An LLM takes the user's intention or mood as input and generates a structured, calming meditation script tailored to their needs.

2. Voiceover Synthesis

The generated script is passed to a text-to-speech engine to produce a natural, soothing audio narration in real time.

3. Visual Generation

An AI image generator creates unique, serene ambient visuals that complement the meditation theme and mood.

4. Session Assembly

The audio narration and generated visuals are combined into a seamless, immersive meditation session presented through a clean web interface.

Technologies Used

React
LLM
Text-to-Speech
Image Generation
Cloud Run

Key Learnings

  • Orchestrating multiple generative AI modalities (text, audio, image) in a single cohesive pipeline.
  • Designing prompt templates that reliably produce calming, structured meditation scripts.
  • Synchronising AI-generated audio and visuals for an immersive user experience.
  • Deploying a multi-modal AI application to Google Cloud Run.