Back to portfolio
Case Study · SaaS Platform

TrackGlow

An AI-powered audio mastering platform that lets homestudio producers and rappers get studio-quality masters in minutes — for free. Upload a track, pick an artist preset, download the result.

Screenshots / app walkthrough

The Challenge / why this exists

The Problem

  • Professional audio mastering costs $50-200 per track — out of reach for most independent artists
  • Mastering requires expensive software (iZotope, FabFilter) and years of trained ears
  • Homestudio producers and rappers often release unmastered music, losing competitive quality
  • Existing AI mastering tools charge subscription fees and offer limited presets

The Solution

  • Upload any .mp3 or .wav file (beat or full track) and get a mastered version back in minutes
  • Choose from 11 artist presets (Drake, Travis Scott, The Weeknd...) or 4 style presets
  • AI-powered 2-stage pipeline: EQ/loudness matching + artist-specific character effects
  • Free to use — built entirely on free-tier cloud infrastructure

Tech Stack / what powers it

A full-stack TypeScript + Python platform spanning 6 different cloud services — all on free tiers. The frontend runs on Vercel, the audio worker on Fly.io, with Supabase, R2, and Redis gluing it all together.

Frontend

Next.js 16 React 19 TypeScript Tailwind CSS v4

Auth

NextAuth v5 Google OAuth JWT Sessions

Database

Supabase PostgreSQL Prisma v6

Storage & Queue

Cloudflare R2 Upstash Redis

Audio Engine

Python 3.11 matchering Spotify pedalboard FFmpeg

Hosting

Vercel Fly.io Resend
Monthly Infrastructure Cost
$0
6 free-tier services: Vercel + Supabase + Cloudflare R2 + Upstash Redis + Fly.io + Resend

Key Features / what it does

11 Artist Presets

Drake, Travis Scott, The Weeknd, Tupac, Lil Wayne, Wu-Tang Clan, Playboi Carti, Don Toliver, 6ix9ine, Griselda Records, and Vybz Kartel. Each with custom EQ profiles and effects chains.

2-Stage Audio Pipeline

Stage 1: Spectral EQ matching and loudness targeting using pre-computed .npz profiles. Stage 2: Artist-specific pedalboard effects (compression, saturation, stereo widening, sidechain pump).

Intensity Control

A 0-100% wet/dry slider lets users dial in exactly how much of the mastering effect they want — from subtle to aggressive.

Version Retry System

Try up to 3 different presets on the same upload, compare the results side-by-side with audio previews, then download your favorite version.

BPM & Key Detection

The Python worker automatically detects tempo (BPM) and musical key during processing. Results are stored in the database and displayed on the status page.

Custom Reference Upload

Don't want a preset? Upload your own reference track and the AI will match its spectral profile and loudness characteristics.

Admin Dashboard

Full admin panel with user management, job monitoring, preset configuration, feature toggles, and site-wide settings.

SEO & Blog

JSON-LD structured data (SoftwareApplication + FAQPage), dynamic sitemap, Open Graph tags, and a built-in blog with admin CRUD and scheduled publishing.

GDPR Compliant

Cookie consent banner that gates Google Analytics 4 tracking. GA4 events fire on mastering job completion for conversion tracking.

Architecture / how it works

A decoupled architecture with the Next.js frontend on Vercel and a Python audio processing worker on Fly.io, communicating through Redis queues and S3-compatible storage.

Processing Pipeline

Upload Presigned PUT to R2
Next.js API Create job in DB
Redis Queue LPUSH job ID
Python Worker matchering + pedalboard
R2 Output master.wav stored
Poll & Download Frontend polls status

Like what you see?

I build full-stack SaaS platforms from scratch — Next.js, Python, AI, cloud infrastructure. All running on free tiers. Let's talk about your project.