B-Side
Date
July 2025
Description
An AI music composition co-pilot combining real-time musical input analysis with intelligent vocal interaction.
The system processes live MIDI input through custom detection algorithms, feeds musical context and voice commands to OpenAI for intelligent responses, and uses Google Magenta for automated arrangements, creating an interactive environment that understands both your playing and spoken requests.
Arrangement Generation Demo
Real-time MIDI analysis detects chord progressions and generates AI-powered bass lines and drum patterns using Google Magenta models.
Creative Advice Demo
Voice-activated AI assistant provides contextual musical guidance by understanding what the user is currently playing and offering personalized suggestions.
Watch the Full Presentation!
Tech Stack
User Interface
Astro
Modern web framework for interactive applications with optimal performance.
Backend/API
FastAPI
High-performance Python framework enabling asynchronous MIDI analysis.
Vite
Lightning-fast build tool with instant hot module replacement.
MIDI Processing
Python
Backend language chosen for machine learning and audio processing capabilities.
Mido
Specialized library for capturing and analyzing MIDI messages.
NumPy
Mathematical library for timing analysis and chord calculations.
AI Models
Web Speech API
Browser-native speech recognition for real-time voice commands.
Magenta
Google's music AI models for arrangement generation.
.png)
OpenAI API
GPT-powered AI providing contextual musical guidance.
Storage (Coming Soon)
SQLite
Lightweight database enabling real-time feedback loops that personalize AI responses as users interact.
User
Speech Processing
MIDI/Audio Interface
OpenAI Integration
MIDI Analysis Engine
Arrangement Generation
Backend Infrastructure
System Architecture Flow
Presentation Slides
