DirectAI Music Platform

AI-Powered Music Analysis & Generation Platform - Production-ready platform for music studios, producers, and enterprises.

🚀 Status: MVP Complete ✅ | Ready for Azure Deployment

A comprehensive music analysis and generation platform leveraging state-of-the-art AI models (Demucs, MusicGen) to analyze audio structure, separate stems, extract musical features, and generate new music programmatically.

📘 Deployment Plan | 🚀 Quick Start Guide | 📋 MVP Status

✨ Core Features

Audio Analysis

• Stem Separation: Demucs v4 (htdemucs) - 4-stem isolation (vocals, drums, bass, other)
• Musical Analysis: BPM detection, key/scale detection, chord extraction, beat tracking
• Metadata Extraction: ID3 tags, album artwork, duration, format
• Real-Time Updates: SignalR WebSocket notifications for job progress

AI Music Generation (MusicGen)

• Model: Meta's MusicGen-small (300MB, 8-second load time)
• Text-to-Music: Generate music from text prompts
• Controllable Parameters: BPM, key, style, duration
• Training Pipeline: Fine-tune on custom stems for personalized generation
• Output: 32kHz generation, resampled to 44.1kHz WAV

Production Features

• Async Job Processing: Service Bus queues with auto-scaling workers
• Blob Storage: Azure Storage for audio files, stems, and training data
• Database: PostgreSQL 16 with Entity Framework Core
• API: .NET 8.0 REST API with Swagger documentation
• Cross-Platform UI: .NET MAUI desktop app (Windows/Mac)

🏗️ Architecture

Technology Stack

• .NET 8.0: Web API, Entity Framework, SignalR, MAUI
• Python Workers: FastAPI microservices (analysis + generation)
• AI Models: Demucs v4, MusicGen, librosa, essentia, madmom
• Azure: Container Apps, PostgreSQL, Blob Storage, Service Bus
• Containerization: Docker Compose (local), Azure Container Apps (production)

Local Development

┌─────────────┐     ┌──────────────┐     ┌─────────────────┐
│   .NET API  │────▶│  PostgreSQL  │     │  Blob Storage   │
│  (Port 5000)│     │  (Port 5432) │     │   (Azurite)     │
└──────┬──────┘     └──────────────┘     └────────┬────────┘
       │                                           │
       ├───────────┬───────────────────────────────┤
       │           │                               │
       ▼           ▼                               ▼
┌─────────────┐ ┌─────────────┐          ┌─────────────┐
│  Analysis   │ │ Generation  │          │   Service   │
│   Worker    │ │   Worker    │          │     Bus     │
│ (Port 8001) │ │ (Port 8080) │          │   Queues    │
└─────────────┘ └─────────────┘          └─────────────┘

Azure Production (Per Customer)

Customer Subscription
├── Container Apps Environment
│   ├── API (1-10 replicas, always-on)
│   ├── Analysis Worker (0-5 replicas, auto-scale)
│   └── Generation Worker (0-3 replicas, auto-scale)
├── PostgreSQL Flexible Server (B2s)
├── Blob Storage Account (LRS/GRS)
├── Service Bus Namespace (Standard)
├── Container Registry
├── Application Insights + Log Analytics
├── Key Vault (secrets management)
└── Managed Identity (RBAC security)

🚀 Quick Start

Prerequisites

• Docker Desktop
• .NET 8.0 SDK
• Git

Run Locally

# Clone repository
git clone https://github.com/TheManInTheBox/DirectAI.git
cd DirectAI

# Start all services
docker compose up -d

# Check service health
docker compose ps
docker compose logs api --tail=20

# Access services
# API: http://localhost:5000
# Swagger: http://localhost:5000/swagger
# PgAdmin: http://localhost:5050

Deploy to Customer Azure Subscription

# Set customer context
azd env new customer-acme-music
azd env set CUSTOMER_NAME "acme-music"
azd env set AZURE_SUBSCRIPTION_ID "customer-sub-id"

# Deploy infrastructure + apps
azd up --subscription "customer-sub-id"

# Verify deployment
azd show
azd monitor --logs

📘 Full Deployment Guide

📊 Current Status (MVP Complete)

✅ Backend Infrastructure

• .NET 8.0 API with async job processing
• PostgreSQL database with EF Core migrations
• Blob storage integration (Azurite local, Azure production)
• SignalR real-time job updates

✅ Analysis Worker

• Demucs v4 stem separation (~30s per track)
• BPM detection, key/scale analysis
• Chord extraction with timeout handling
• Beat tracking and metadata extraction
• MP3 album artwork support

✅ Generation Worker

• MusicGen-small integration (8s startup)
• Text-to-music generation
• 32kHz → 44.1kHz resampling
• Compatible with analysis worker output

✅ Deployment Ready

• Docker Compose for local development
• Azure Container Apps deployment plan
• Bicep infrastructure-as-code templates
• Per-customer isolation architecture

🔄 In Progress

• Training pipeline for MusicGen fine-tuning
• Stem selection UI for dataset creation
• Parameter-controlled generation (key, BPM, bars)
• Customer onboarding automation

📋 Detailed MVP Status | 🧪 Testing Guide

📚 Documentation

• Get Started Guide - Local setup and first analysis job
• MVP Status - Current implementation status and roadmap
• Testing Guide - How to test all features
• Deployment Plan - Azure Container Apps deployment strategy
• Architecture Overview - System design and data flow
• API Reference - REST endpoints documentation
• Worker Documentation - Analysis & generation workers

🎓 Key Capabilities

For Music Producers

• Upload tracks → Get separated stems in 30 seconds
• Analyze chord progressions and key signatures
• Extract BPM and beat grids for remixing
• Generate new stems matching your musical style

For Developers

• RESTful API with OpenAPI/Swagger
• Real-time WebSocket updates via SignalR
• Async job processing with retry logic
• Comprehensive logging and telemetry

For Enterprises

• Isolated Azure deployment per customer
• Managed identity security (no credentials)
• Auto-scaling workers (cost-efficient)
• Centralized monitoring via Azure Lighthouse

️ Development

Project Structure

DirectAI/
├── src/
│   ├── MusicPlatform.Api/          # .NET 8.0 REST API
│   ├── MusicPlatform.Domain/        # Core business logic
│   └── MusicPlatform.Infrastructure/ # Data access, external services
├── workers/
│   ├── analysis/                    # Python FastAPI (Demucs, analysis)
│   └── generation/                  # Python FastAPI (MusicGen)
├── infrastructure/
│   ├── main.bicep                   # Azure IaC templates
│   └── kubernetes/                  # K8s manifests (optional)
├── database/
│   └── schema.sql                   # PostgreSQL schema
├── .azure/
│   └── plan.copilotmd              # Deployment plan
└── docker-compose.yml              # Local development

Running Tests

# .NET API tests
dotnet test tests/DirectML.AI.Tests/

# Test analysis worker locally
docker compose up -d analysis-worker
curl http://localhost:8001/health

# Test generation worker locally
docker compose up -d generation-worker
curl http://localhost:8080/health

Building for Production

# Build all Docker images
docker compose build

# Push to Azure Container Registry
az acr login --name <registry-name>
docker tag directai-api <registry>.azurecr.io/directai-api:latest
docker push <registry>.azurecr.io/directai-api:latest

🤝 Contributing

This is a commercial product under active development. For partnership opportunities, licensing, or technical inquiries:

• GitHub Issues: Bug reports and feature requests
• Discussions: Architecture and design conversations
• Pull Requests: Welcome for bug fixes (requires CLA)

📄 License

Proprietary - All rights reserved. Commercial licensing available.

Contact: GitHub Profile

---

Built with: .NET 8.0 • Python 3.10 • Azure • Docker • PostgreSQL • MusicGen • Demucs

Powered by: Meta AI (MusicGen) • FAIR (Demucs) • MTG (Essentia) • madmom