Why
I built this app out of own experience: Finding the perfect background music to focus takes valuable time away from the actual work. Browsing several playlists to find that one track that fits your mood is tedious.
What
I started Lucid Tone and built the real-time audio system behind it. Rather than serving another playlist, each session composes an evolving stream with fine-grained controls that shape the music around the listener’s mood.
How
A FastAPI server runs a Markov chain–based music engine and streams each composition to the client over WebSockets. React and TypeScript power the listening interface, while Redis maintains session state and supports stream continuity across persistent connections.
Compose instead of catalog
The generative engine is the product. A larger fixed library would only spread repetition across more tracks, eventually making the listening experience predictable and stale.
Give attention an arc
Entry, anchor, sustain, and refocus phases give each session direction, allowing the music to support changing attention without requiring the listener to manage it.
Treat egress as a constraint
Minimizing audio egress was a core engineering priority. The streaming format and delivery strategy were designed around the right balance of sound quality, performance, cost, and connection stability.